Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

Re-evaluation of fatigue curves for flush ground girth welds (October 2008)

Y-H Zhang and S J Maddox

TWI, Cambridge, United Kingdom

A Stacey
HSE, London, United Kingdom

Paper presented at 12th International Symposium on Tubular Structures (ISTS 12), Shanghai, China, 8-10 October 2008.


The fatigue performance of girth welded steel pipes with the weld cap and root flush-ground is designated as Class C in the BS 7608 fatigue design rules. However, this is not based directly on experimental data. Since fatigue test results for flush-ground girth welds in steel pipes have now become available, a comprehensive examination of published work was carried out to re-evaluate this and other design S-N curves for such welds on the basis of the relevant experimental data. The data considered included some from tests on full-scale welded pipes and other from strip specimens cut from girth-welded pipes. In addition, the opportunity was taken to consider the larger database available from butt-welded plate specimens for comparison.

An important condition on the use of relatively high design curves for flush-ground butt welds is that the weld should be proved free from significant welding defects. Concern about the detectability of small defects that could result in fatigue performance below such design curves has limited their acceptance. In the case of volumetric flaws, relevant fatigue data obtained from girth welds containing reportable embedded flaws are available and they were used, together with data obtained from butt-welded plate specimens, to address such concerns.

1 Introduction

Flush-grinding of a butt weld is an established method for improving its fatigue performance. This eliminates the stress concentration created by the weld profile and removes the inherent weld toe flaws at which fatigue cracks typically initiate. Consequently a fatigue performance far superior to that typically assigned to conventional structural welds in the as-welded condition is expected. However, the current fatigue design rules in various standards and codes for such welds are not based on fatigue test data from flush-ground girth welds but on data for joints between flat plates, much of which were obtained many years ago. Furthermore, the welding procedures, types and consumables might not be representative of those currently used in offshore structures. However, since data from flush-ground girth welds in steel pipes have now become available, there is the opportunity to re-evaluate the design S-N curve(s) for such welds on the basis of relevant experimental data.

To qualify for [BS 7608 Class C (1993)], flush-ground welds must be free from defects that could reduce that potentially high fatigue strength. However, in the great majority of situations it is found that a fatigue strength higher than Class D cannot be justified. This is attributed mainly, from evidence from plate specimen tests, to the possible presence of flaws which are too small for reliable detection using current non-destructive testing (NDT) methods but which could be of sufficient size to reduce the fatigue strength of the joint. Therefore, it is necessary to examine the defect acceptance criteria for flush-ground welds using data from girth welds, particularly for embedded defects since they become the weak link in the fatigue performance and the minimum defect sizes need to be assessed against the detection capability of current NDT methods.

2 Flush-ground welds with no reportable flaws

2.1 Full scale pipe specimen

Published data from fatigue tests on full-scale pipes with flush-ground girth welds are very scarce and only four relevant test series were found.[Wirsching et al. 1995, Salama 1999, Maddox et al. 2002, Wastberg and Salama 2007] Fatigue test results for seven API-5L X60 UOE steel pipe specimens with outer diameter (OD) 610mm and wall thickness (WT) 20mm were reported by [Wirsching et al. (1995)]. Details of the test results, NDT examination records and failure locations were not reported and only the statistical analysis of fatigue endurances was published.

The results reported by [Wastberg and Salama (2007)] were also from UOE X60 pipe specimens with an OD of 610mm and WT of 20.6mm. They included the data previously published by [Salama (1999)]. All ground welds were inspected by two independent organizations and were found to be either free from any defects or to contain only small defects of acceptable sizes according to [API 1104 (2005)]. Unfortunately, the NDT results were not reported. A total of eight pipe specimens were tested, each containing three girth welds. All fatigue tests were carried out under axial loading at a stress ratio R=0.1. It was found through post-test examinations that fatigue crack had occurred at lack of fusion defects, located close to the inside or outside surface. The defect sizes, which caused significant reduction in fatigue lives, were around 2x15mm and were considered to be at the limits of reliable detection by NDT.

[Maddox et al (2002)] reported results obtained from flush ground girth welds in 609mm OD and 21.4mm WT API 5L X60 steel pipe for tendons and 273mm OD and 12.6mm WT API 5L X80 pipe for risers. Two tendon specimens containing three welds each were fabricated in the 1G position, using single sided submerged arc welding (SAW) throughout. The weld root was made onto backing tape. Three riser specimens containing two welds each were fabricated in the 1G position, using a single sided GTAW root pass followed by SAW fill and cap. All weld root beads and caps were ground flush with the pipe surfaces after welding. Comprehensive NDT including radiography testing (RT), ultrasonic testing (UT) and magnetic particle inspection (MPI) did not reveal any defects. Six full-scale girth welds in the two tendon specimens and six full-scale girth welds in the three riser specimens were fatigue tested under tension-tension axial loading at a mean stress of ~ 175 MPa and 125 MPa, respectively. All the tests gave run-outs with no evidence of fatigue cracking in any of the welds.

These results, or just the reported S-N curves in the case of [Wirsching et al], are plotted in Figure 1 together with the BS 7608 Class C design curve. As will be seen, the fatigue test results for full-scale specimens do not agree well. Although the mean curve from [Wirsching et al. (1995)] was reported to be slightly above the BS 7608 C design curve, the design curve derived is below it. This was due to the large standard deviation in the original data and the small number of results. For the test results reported by [Wastberg and Salama (2007)], the endurances of four of the eight tests were below the BS 7608 design C curve. Failure of these girth welds started from planar defects of about 2x15mm. They were missed during NDT by two independent organisations. For the four pipe specimens which achieved the Class C curve, failure started from defects of ≤1x8mm, which is considered to be too small to be reliably detected. On the other hand, all the results from Maddox et al were run-outs to endurances that exceeded the Class C curve significantly. This highlights the importance of careful NDT and stringent requirement on the elimination of defects to qualify for the Class C curve. For this reason, the test results from [Wirsching et al] should be taken with caution. The work reported by [Wastberg and Salama] raised a question about whether or not a lack of fusion defect, which could significantly reduce the fatigue lives of ground welds, can be reliably detected by relevant NDT.

2.2 Strip specimen

In the early 1990's, a large testing programme [Razmjoo et al. 1998] was carried out in three laboratories to investigate the fatigue performance of strip specimens extracted from flush-ground girth welds in pipes for applications in the Heidrun TLP. The tendons were made from 12m lengths of 1118mm OD by 38mm WT pipes in steel equivalent to API 5L Grade X70 specification. They were fabricated using a number of welding procedures.



Fig.1. Comparison of fatigue endurance of full-scale specimens with the BS 7608 C curve. All specimens tested by [Maddox et al (2002)] were run-outs


Waisted fatigue test specimens with a minimum width of around 100mm were extracted from the pipes. All welds were flush-ground both inside and outside before fatigue testing. The fatigue tests were conducted under constant amplitude axial loading at R=0.1 in most cases. The tests were run until failure occurred or to target lives based on the BS 7608 Class C curve.

From the total of 68 specimens, 18 specimens were tested at 6°C in seawater with cathodic protection (-1050mV). Others were tested at room temperature in air.

In general, establishing the fatigue strengths of the flush-ground butt welds was a challenge. In the absence of weld toes and any significant embedded flaws, fatigue cracking could initiate at various locations in the specimens other than the weld. In the event, a test was terminated for one of four reasons:

  1. Failure in the weld (11 specimens)
  2. Failure in the weld but from the edge of the specimen (11 specimens)
  3. Failure from the machine grips or in parent plate (22 specimens)
  4. Run-out (i.e. specimen did not fail) (24 speci-mens)

The test results are compared with the Class C curve in Figure 2. It can be seen that all the test data are above the C curve, including those from tests carried out in seawater cathodic protection (CP) [Razmjoo et al. 1998]. Furthermore, these data from strip specimens exhibit a much shallower slope in the long life regime (>106 cycles) when compared with the Class C curve, and a higher fatigue limit than that of the full-scale pipes shown in Figure 1.



Fig.2. Comparison of the fatigue performance of flush-ground strip specimens cut from girth welds [Razmjoo et al. 1998] with the BS 7608 C curve.


2.3 Plate specimen

As the S-N curve classification of welded plates is the same as that for girth welded pipes, the review of the fatigue performance of flush-ground butt welded steel plate specimens conducted by [Maddox (1997)] is useful, see Figure 3. Data in the low-cycle fatigue regime were expressed in terms of the pseudo-elastic stress range, as suggested in [PD 5500 (2000)]. As will be seen in Figure 3, the plate specimen data are more widely scattered than the girth weld strip specimen data. The majority of the results qualified for the C curve with only a few exceptions where the results were only slightly below the C curve. The larger scatter is to be expected since the database is much larger than that for strip specimens. However, it might also be associated with inconsistent welding quality in the plate specimens obtained from many different sources, in contrast with the comparatively good welding control and stringent NDT acceptance criteria in girth welding. Consequently, some of the plate specimens might have contained defects larger than the acceptable limit for girth-welded pipes. Furthermore, the tests on these plate specimens were carried out at different stress ratios, some even with R<0. Overall, the results from plate specimens suggest that the C curve is applicable to flush ground butt welds.


Fig.3. Fatigue endurance data for flush-ground butt-welded plate specimens [Maddox 1997]



Fig.4. Comparison of fatigue design curves for flush ground butt welds from different standards


2.4 Comparison of fatigue design rules for flush ground welds

Generally, flush-ground butt welds are recognised as having enhanced fatigue performance compared with as-welded joints. This is reflected in the different design standards. A graphic presentation of the corresponding design S-N curves is shown in Figure 4, where it can be seen that the Class C curve in BS 7608 has a slope of 3.5, whereas 3.0 is adopted by [DNV (2005)] and IIW [Hobbacher 1996]. In comparison, Class C is more conservative for endurances below 2x105 cycles but less conservative above this endurance. The AWS designation for flush-ground butt welds is Class B which has a shallow S-N curve (m=4.3). As noted earlier, a shallow slope with m>3 would be expected to be appropriate for flush-ground butt welds as a result of crack initiation occupying a significant proportion of the fatigue life. Compared to other standards, the AWS B curve is conservative at fatigue endurances below one million cycles, especially in the regime of high stress ranges. It should be noted that the AWS C1 curve has been reported to be more appropriate as the design curve for flush-ground girth welds. [Buitrago & Zettlemoyer 1999]

It must be emphasized that the above comparison is based on the S-N curves at the reference thickness for each code. Since some codes apply a thickness correction for flush ground welds for thickness above the reference value and recommend some degree of penalty, care must be taken when such a comparison is made at a specimen thickness beyond the reference thickness for a certain code.

3 Flush ground girth welds with reportable flaws

3.1 Surface breaking and internal planar defects

There do not appear to have been any investigation specifically directed at studying the effect of surface breaking or internal planar defects, revealed during NDT, on fatigue performance of flush ground welds. The lack of fusion defects reported by [Wastberg and Salama (2007)] were only revealed after the post test examinations of the fracture surfaces of the failed welds. In fact, to qualify for the fatigue design curves for flush-ground girth welds, the criteria of defect acceptance are very strict in relevant standards. [BS 7910 2005, ASME 2007] Surface breaking and internal planar defects, such as lack of fusion or lack of penetration, are particularly important since they are much more deleterious to the fatigue performance of the weld than embedded volumetric defects such as pores or slag inclusions. In recognition of this, it has been proposed that they cannot be accepted in flush-ground welds in TLP tendons designed to AWS C1.[Buitrago & Zettlemoyer 1999]

3.2 Effect of porosity

Recently the effect on fatigue performance of porosity and slag inclusions in flush-ground girth welds has been investigated [Razmjoo et al. 1998, Buitrago & Zettlemoyer 1999]. Figure 5 shows the fatigue test results from the strip specimens containing porosity. The maximum pore sizes have been characterised by RT and UT in these specimens. As it is likely that fatigue crack initiation may occur from internal flaws, the fatigue performance of the plate specimens (12mm in thickness) containing porosity density <4% [Harrison 1972a] are also included for comparison. It will be seen that when porosity density was less than 4% and the maximum reported individual flaw size was up to 4.8mm, the fatigue performance could still be qualified as Class C, except in the high stress range/low endurance (<105 cycles) regime. The results suggest that, in the medium and long fatigue life regimes, a pore size of around 4mm can be tolerated, a size which can be detected reliably by RT or UT.

However, when porosity density was increased to 8% the test data for many plate specimens fell below the C curve, Figure 6 [Harrison 1972a]. This suggests that when pore density is low (below 4%), the individual pore size plays an important role. When the pore density is high, however, the fatigue performance is significantly reduced regardless of the individual pore size.


Fig.5. Fatigue performance of strip and plate specimens containing porosity less than 4%


Fig.6. Fatigue performance of plate specimens containing porosity less than 8% [Harrison 1972a]

3.3 Effect of slag inclusions

Fatigue test results from flush-ground girth welds in strip specimens with slag inclusions [Razmjoo et al. 1998, Buitrago & Zettlemoyer 1999] are compared in Figure 7. To illustrate the effect of inclusions, the fatigue data from as-welded plates containing inclusions less than 10mm long [Harrison 1972b] were also included for comparison. It can be seen that, although the data from [Razmjoo et al. (1998)], which had defect lengths between those reported in two other references [Buitrago and Zettlemoyer 1999, Harrison 1972b], qualify for Class C, many of the test results from [Buitrago & Zettlemoyer and Harrison] fall below the C curve even from the plate specimens. This suggests that the fatigue life de-pends not only on the inclusion length, as specified in [BS 7910 (2005)], but also on other factors such as the ligament and the inclusion height.


Fig.7. Fatigue performance of strip and plate specimens containing slag inclusions.

4 Discussion

When considering the fatigue design of girth weld in pipes or tubes, results obtained from full-scale specimens are highly recommended in order to exclude uncertainties associated with other types of specimen. However, the requirements for the special testing facility, time and cost make this difficult to achieve. In practice, small-scale specimens have often been used to simulate the behaviour of large-scale components. When the test results from these specimens are used for design purposes, however, care must be taken to ensure they represent the fatigue performance of the large-scale components.

Available fatigue data from flush-ground welds in strip specimens cut from girth welded pipes confirm the qualification of the BS 7608 C curve. Even those specimens that were originally rejected due to the presence of unacceptable defects [Razmjoo et al. 1998] gave lives consistent with Class C. This suggests that the fatigue performance of strip specimens free from defects can be comfortably classified as Class C. However, some of the results of the flush-ground girth welded pipes were lower than the corresponding strip specimen results and could not be qualified for the C curve [Wirsching et al. 1995, Wastberg and Salama 2007]. Although the data from the above full-scale specimens should be viewed with caution, as described before, the difference in fatigue performance between the strip and the full-scale specimens was not unexpected as the same has been observed in tests on as-welded girth welds.[Razmjoo et al 1998] In this case, the difference was attributed to the possible effects of size and residual stress.

With regard to size, as a strip specimen contains only a small proportion of the length of a girth weld, it is unlikely to contain the most severe defect present or to introduce the same level of stress concentration as in some locations in the pipe. With respect to residual stress effects, it is speculated that the welding-induced residual tensile stress could be partly or even fully released during extraction of a strip specimen. Even if this were not the case, residual stress measurements in girth welds have shown that they can be widely scattered [Maddox et al. 2006], even for a single weld. Thus, a strip specimen may coincide with a region of low residual stress in the original girth weld. A consequence of having lower residual stresses in the strip specimens is that the effective mean stress, resulting from the super-position of the applied and residual stress, would have been lower than in the full-scale specimens.

Another issue, which could affect the direct comparison of the fatigue results between strip and full-scale specimens, is the quality of flush-grinding at weld roots. For single sided welds in pipes, grinding the weld roots is not as easy as for strip specimens. Consequently cracking might preferentially initiate at the weld root, resulting in a lower fatigue endurance. Thus, to make the fatigue results of strip specimens representative of large-scale specimens, the same quality of grinding both inside and outside must be assured.

To qualify for the Class C design curve, detailed NDT of welds must be conducted to ensure that they are free from significant defects. Thus, any test data for which the NDT results are lacking, e.g. the data reported by [Wirsching et al] from full-scale fatigue testing, should be taken with caution.

The work carried out by [Maddox et al (2002)] provided full details of the NDT examination and testing conditions. All twelve welds qualified with respect to the C curve without failure. The results therefore strongly support the BS 7608 C curve. Furthermore, all the results from the strip specimens also support the adoption of the C curve for flush-ground girth welds, even for those tests undertaken in seawater with cathodic protection.[Razmjoo et al. 1998]

The defect acceptance criteria in two codes, which might be used for flush-ground girth welds, are compared in Table 1. One is based on fitness-for-service [BS 7910 2005] and the other on fabrication limits (ASME 2007). It should be noted that BS 7910 only provides defect limits up to quality category Q1 (equivalent to design Class D) on the basis that beyond this limit NDT cannot be relied upon to detect critical defects. The defect acceptance criteria in ASME, Section VIII, Division 1 are those currently used for some tendons.[Buitrago & Zettlemoyer 1999]

The results obtained from strip specimens cut from girth welded tendons reported by [Razmjoo et al (1998)] provided a direct comparison between specimens containing reportable defects, some of which were measured, and others with no reportable defects under the same production procedures. A pore size up to 8mm and slag inclusions up to 18mm long were reported. The comparable fatigue endurances of these specimens containing reportable defects with those containing no reportable defects suggest that these defects can be tolerated, providing confidence in detecting the limiting defect sizes using current NDT methods. However, this finding was based on a comparatively small sample and more such tests are required to determine the critical defect sizes.

Table 1 Defect acceptance criteria for quality category Q1 [BS 7910 2005] and flush-ground girth welds with thickness limited to 50mm.[ASME 2007]

*: can be calculated by fracture mechanics but considered to be undetectable by current NDT methods.

Flaw acceptance limit
StandardPorosity, % area on radiographIndividual pore size, mmSlag inclusion length, mmUndercut, (depth/wall thickness)Planar defects
BS 7910-Q1 3.0 thickness/4 or 6.0 2.5 0.025 not applicable*
ASME VIII Div.1 not defined 6.3 6.0 no undercut not allowed

5 Conclusions

  • There is uncertainty in the fatigue performance of flush ground girth welds in full-scale pipe specimens. Although most of the results reviewed strongly support the BS 7608 C curve, some data, not reported in full (lacking details of NDT records, failure locations and individual test results) suggest that Class C can be unsafe. They highlight the importance of careful NDT inspection and stringent requirement on the elimination of defects to qualify for the ClassC curve. It is apparent that there is a need for further effort to establish the missing details and for more full-scale fatigue testing of flush-ground girth welds with full details of the welds.
  • BS 7608 Class C is comfortably qualified on the basis of fatigue data obtained from strip specimens cut from girth welded joints.
  • The optimistic fatigue performance of small specimens can be attributed partly to their reduced residual stresses compared to large-scale pipe specimens. Fatigue testing of such specimens at high stress ratios is required in order to predict the behaviour of actual girth welds conservatively.
  • Data from flush-ground plate specimens also support Class C classification. However, because of deficiencies, for example, NDT records, weld quality and misalignment, their fatigue behaviour may not be truly representative of that of girth welds.
  • The limited database on flush-ground girth welds containing reportable embedded flaws suggests that the flaw sizes present in welds which achieved Class C can be detected using current NDT methods.

6 References

API 1104. Welding of Pipelines and Related Facilities, American Petroleum Institute, November 2005.

ASME Section VIII, Division 1. 2007. Boiler and Pressure Vessels and Code'. The American Society of Mechanical Engineers, New York.

AWS D1.1/D1.1M:2002: 'Structural Welding Code - Steel', American Welding Institute, August 31 2001.

BS 7608:1993. Fatigue design and assessment of steel structures, BSI, London, UK.

BS 7910:2005. Guide on methods for assessing the acceptability of flaws in metallic structures. BSI, London, UK.

Buitrago J, Weir M S & Kan W C. Fatigue design and performance verification of deepwater risers. In Proc. of OMAE2003, Paper No. OMAE 2003-37492, Cancum, Mexico, June 2003.

Buitrago J & Zettlemoyer N. 1999. Fatigue of tendon welds with internal defects. In Proc. of OMAE99, Paper No.: Mat-2001, St. Johns, Newfoundland, Canada.

DNV RP C203:2005: 'Fatigue strength analysis', Det Norske Veritas, May 2005.

Harrison J D. 1972a. The Basis for An Acceptance Standard for Weld Defects, Part 1: Porosity. Metal Construction, Vol. 4, pp.99.

Harrison J D. 1972b. The basis for an acceptance standard for weld defects, Part 2: Slag inclusions. Metal Construction, 4, pp.262.

Hobbacher A: 'Fatigue design of welded joints and components', Recommendations of IIW Joint Working Group XIII-XV, Abington Publishing, 1996.

Maddox S J. 1997. Developments in fatigue design codes and fitness-for-service assessment methods. In Performance of Dynamically Loaded Welded Structures, IIW 50th Annual Assembly Conference, San Francisco, Ed: S J Maddox, M Prager. Publ: New York, NY 10017, USA; Welding Re-search Council, Inc., pp.22-42.

Maddox S J, Speck J B, Lockyer S A & Razmjoo G R. 2002. Fatigue performance of girth welds made from one side. TWI Report No. 5680/26/02, confidential to membership companies.

Maddox S J, Razmjoo G R & Speck J B. 2006. An investigation of the fatigue performance of riser girth welds. In Proceedings of Conf. Offshore Mechanics and Arctic Engineering, ASME, Paper No. OMAE2006-92315.

PD 5500:2000: 'Unfired fusion welded pressure vessels', BSI, January 2000.

Razmjoo G R, Maddox S J & Hayes B. 1998. Fatigue performance of flush-ground TLP tendon girth welds. TWI Report No. 5680/13/98, confidential to membership companies.

Salama M M. 1999. Fatigue design of girth welded pipes and the validity of using strips. In Proc. of OMAE99, Paper No.: Mat-2003, St. Johns, Newfoundland, Canada.

Wirsching P, Karsan D I & Hanna S Y. 1995. Fatigue/fracture reliability and maintainability analysis of the Heidrum TLP tether system. In Proc. of OMAE95, pp.187, Copenhagen, Denmark.

Wastberg S & Salama M M. 2007. Fatigue testing and analysis of full scale girth welded tubulars. In Proc. of OMAE07, Paper No.: OMAE2007-29399, June 10-15 2007, USA.

For more information please email: