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

Subscribe >
Skip to content

Review and assessment of fatigue data for offshore structural components containing through-thickness cracks (June 2008)

   

Y H Zhang, TWI Ltd, Abington, Cambridge, UK

A Stacey, Health and Safety Executive, London, UK

Paper presented at OMAE 2008 27th international Conference on Offshore Mechanics and Arctic Engineering, Estoril, Portugal, 15 - 20 June 2008. Paper OMAE 2008 - 57503.

Abstract

In recent years, structural integrity management schemes for offshore installations have placed increased reliance on the use of flooded member detection (FMD) as the principal inspection method. This method can be routinely employed in a remotely operated vehicle, which enables a large number of members to be inspected fairly quickly at a much reduced cost compared to using diver operated techniques. However, reliance on FMD for safety assurance requires that welded joints retain sufficient fatigue life and static strength after through-thickness cracking.

A comprehensive examination of published work containing data on fatigue lives beyond through-thickness cracking in offshore structures was carried out, resulting in the development of a database of 281 relevant tests. The database was used to perform a statistical assessment of the effects of different testing conditions and geometrical parameters on the remaining fatigue life beyond the occurrence of through-thickness cracking, N3, which was represented by a parameter Re. Whilst the data showed a large amount of scatter, it was found that Re depends strongly on chord thickness, loading mode, type of joint and testing environment. In some cases, a significant amount of remaining life existed. This was often associated with T-type tubular joints with thin chord thickness under out-of-plane loading and a seawater (with CP) environment. The influence of the relevant parameters on Re is discussed and attributed to their effect on crack shape, stress distribution, cracking location and crack propagation path.

Introduction

Maintenance of structural integrity in offshore installations is of prime consideration for safe operation. One particular concern with structural integrity is associated with cracking from inherent fabrication defects in offshore structures due to cyclic loading. To ensure safe operation of offshore structures, in addition to good design and careful control of fabrication defects, implementation of comprehensive inspection programmes is required. Current practice relies heavily on the use of flooded member detection (FMD) to maintain structural integrity.

There has been an increasing trend towards the use of FMD for the detection of through thickness cracks, as confidence in the ability to predict the performance of offshore structures has grown.[1] FMD is particularly effective for certain situations which are difficult to control at the design and fabrication stages and for redundant members. Furthermore, it is the only technique routinely employed by remotely operated vehicles (ROV), thus providing a good tool for rapid screening of platform members for gross damage and enabling a large number of members to be inspected fairly quickly at a much reduced cost compared to using diver operated techniques. General visual inspection and FMD have accounted for the detection of the highest number of members with significant cracking. In this respect, the use of FMD techniques has grown rapidly in the North Sea and some operators now depend principally on the use of this method in conjunction with visual inspection for the maintenance of structural integrity offshore. In future it is thought that ROVs will perform all of the subsea tasks and an even greater reduction in inspection will result.[2]

Fatigue design in offshore structures is based on S-N curves derived from tests on tubular joints where failure is defined as penetration of wall thickness.[3,4] However, in practice, fatigue cracks are likely to continue to grow around the weld circumference after breaking through the wall and it has been found in certain circumstance that a significant residual life remained after through-thickness cracking.[5-12] On the other hand, reliance on FMD for safety assurance requires that sufficient fatigue life and strength of tubular joints remain after through-thickness cracking such that remedial actions can be taken. Therefore, a sound knowledge of the remaining life beyond through-thickness cracking is essential for:

  1. safe operation of offshore installations;
  2. inspection planning; and
  3. life extension.

Research in this area has been relatively limited. In a recent comprehensive examination of published works containing data on fatigue lives beyond through-thickness cracking in offshore structures,[13] the effects of many variables on the remaining life, including stress ratio, geometric dimensions, loading mode, joint type, post-weld heat treatment, stiffener type, weld toe profile, testing environment, variable amplitude load, compressive end load on chord, location of failure, girth welds, were examined and assessed. In this paper, the assessment of the effects of some of these parameters on the remaining life are reported.

Nomenclature

a Crack depth
c Crack half-length
COV Coefficient of variation
d Brace diameter
D Chord diameter
K Stress intensity factor
L Chord length
N3 Fatigue life to through-thickness cracking
N4 Total fatigue failure
R R-ratio
Re Remaining fatigue life ratio
SD Standard deviation
t Brace wall thickness
T Chord wall thickness
α Chord length parameter
β Joint diameter ratio
γ Chord diameter to thickness parameter
τ Joint thickness ratio

Assessment methodology

For tubular joints, the number of cycles corresponding to four stages during the fatigue life is usually denoted as N1 to N4:

  • N1 represents the point at which a crack is first noted by any method, or 15% change in strain output from a gauge located close to the point of crack initiation;
  • N2 represents first visual crack detection;
  • N3 represents the attainment of a through-thickness crack;
  • N4 represents the end of test.

Several methods were used to determine N3 in the literature: alternating current potential drop (ACPD), strain drift, visual observation, loss of internal pressure. N3 was sometimes not easily defined precisely, and it is likely that some degree of systematic variation in N3 lives between different investigators for similar testing condition and specimen geometry could be present simply due to the different detection criteria used.

The common criteria used to determine N4 were: limit of actuator stroke, a certain distance of crack propagation away from the weld toe,[5-10] or complete brace separation. Since actuator displacement invariably increased dramatically towards the end of a test, the variations in N4 lives between different investigators were not generally regarded as significant.[5-6]

Residual life of through-thickness cracked members in this study is evaluated in terms of a parameter Re, which is defined as:

spyhzjune08e1.gif
[1]

For each mean Re value determined for each group of data with similar testing conditions and specimen geometry, the associated values of standard deviation (SD) and coefficient of variation (COV) were also determined. COV is a parameter used for assessing data scatter and is defined as the ratio of SD to Re.

As different investigators used different criteria defining N3 and N4, and some obtained remaining lives for through-thickness cracks systematically lower than others even though testing conditions and specimen geometry were similar, the prime consideration in all assessments was, if possible, to analyse data obtained by a single investigator in a single testing programme. If such data were not available or the number of tests was limited for a certain condition, the second priority was given to those sets of data which, except for the parameter to be assessed, had close similarities in testing conditions and specimen geometry.

Review of available fatigue data on N3 & N4

A comprehensive examination of published work containing fatigue lives corresponding to N3 and N4 of tubular structural members was carried out. In total 281 tests with such information were found.[3,5-12,14-23] Most of the data came from the United Kingdom Offshore Steels Research Project (UKOSRP) and large programmes funded by ECSC.[3,14,22] Details of each test, including joint geometry, joint type, loading mode, stress ratio, nominal and hot spot stress ranges, N3 and N4 lives, remaining life parameter Re, detection methods for N3 and N4 lives can be found in[13].

The mean values of Re, SD and COV of the whole database (281 tests) were determined and are given in Table 1. On average, the remaining life of a through-thickness cracked member relative to N3 was about 44%. However, the remaining lives exhibited a large scatter, with an average COV value up to 1.2. The scatter of Re values can also be seen from Figure 1 where the occurrence frequency distribution of Re values is plotted (the low frequency distribution of Re values beyond Re=2.10 is omitted for clarity of the figure). It shows a large scatter in remaining lives beyond N3 with Re values ranging from a minimum of 0.0014[17] to a maximum of 4.95.[12] The cumulative distribution of all Re values is provided in Figure 2. It can be seen that 10% of the tests had Re values below 0.074 and 0.90 of the tests had Re values falling in the range between 0.044 to 1.20.

Table 1: Remaining life for database of through-thickness cracked members

Sample sizeMean ReSDCOV
281 0.443 0.531 1.20
spyhzjune08f1.gif
Fig.1. Probability distribution of remaining life parameter, Re
spyhzjune08f2.gif
Fig.2. Cumulative frequency distribution of Re

Statistical analysis was carried out on N3 and N4 endurance of all data collected and the slopes were forced to be 3.0 in order for comparison with the T' design curve for tubular joints.[4] As most of the tests were conducted in air environment with only fifteen specimens tested under free seawater corrosion and eight specimens under cathodic protection (CP), the HSE T' curve for air is plotted to provide an indication of the relative positions of these N3 and N4 curves. The mean and mean-2SD curves in terms of N3 and N4 are shown in Figure 3. The ratio N4/N3 is 1.38 and 1.25 for the mean and mean-2SD curves, respectively. The scatter of the data is so large that the mean endurance is about one order longer than that for the mean-2SD curve and even the mean-2SD for the N4 endurance falls below the T' design curve. It should be noted that, as there were so many variables in these tests (e.g. weld profile, tubular size, environment, etc.), the mean-2SD curves derived should not be interpreted as being for fatigue design.

Fig.3. Mean and mean-2SD curves in terms of N3 and N4 endurance
Fig.3. Mean and mean-2SD curves in terms of N3 and N4 endurance

Although this scatter was caused, to some extent, by the differences in testing and inspection facilities, and N3 and N4 definition criteria, the analysis suggests that the remaining lives of through-thickness cracks strongly depend on the testing conditions and specimen geometries. This is examined further below.

Assessment of the remaining life of joints with through-thickness cracks

For all data collected various testing conditions and specimen geometries were involved. To facilitate comparisons, a reference condition has been selected, namely T joints in the as-welded condition, tested in air and under constant amplitude loading.

Effect of stress ratio on Re

Stress ratio, R, is defined as the ratio of the minimum stress to the maximum stress in a fatigue cycle. Ten T joints with a chord diameter of 456mm and a chord thickness of 16mm were tested within the ECSC funded programme and results were reported in[3], five at R=0 and five at R=-1, all under axial loading. Comparisons of mean Re values did not suggest a significant difference at the two R ratios, with the mean Re value at R=0 being only slightly higher than that at R=-1, see Table 2. Data enabling such comparison to be made for other types of joints and loading modes were not available.

Table 2: Remaining life statistics for key variables

VariableSample sizeMean ReSDCOVReference
R R=0 5 0.253 0.072 0.28 3
R=-1 5 0.220 0.045 0.21 3
β for T joints tested under OPB loading (D=168mm and β=0.5 6 1.057 0.584 0.55 5
β=1.0 6 1.080 0.353 0.33 5
β for T joints tested under OPB loading (D=457mm and β=0.25 6 0.699 0.591 0.85 6
β=1.0 3 0.498 0.159 0.32 6
τ under axial loading τ=0.55 5 0.400 0.256 0.64 12
τ=1.0 3 0.279 0.168 0.6 12
τ τ=0.6 3 0.498 0.159 0.32 6
τ=1.0 3 0.237 0.105 0.44 6
T (mm)(OPB loading) T=6 12 1.069 0.483 0.45 5
T=16 12 0.533 0.469 0.88 6
Loading mode (T=6mm) Axial 17 0.288 0.218 0.76 12
OPB 12 1.069 0.483 0.45 5
Loading mode (T=16mm) Axial 20 0.289 0.143 0.49 3, 12
OPB 12 0.533 0.469 0.88 6
Loading mode (for K and KT joints with T=16mm) Axial 2 0.125 0.034 0.27 10
OPB 8 0.598 0.180 0.30 10
Axial and IPB loading modes at T=6mm (T joints) Axial 17 0.288 0.218 0.76 12
OPB 14 0.377 0.262 0.69 12
T and K/KT joints with T=6mm T 12 1.069 0.483 0.45 5
K & KT 8 0.378 0.218 0.58 7
T and K/KT joints with T=16mm T 12 0.533 0.469 0.88 6
K & KT 8 0.598 0.180 0.30 10
Environment for K and KT joints under OPB loading Air 8 0.378 0.218 0.58 7
SW (FC) 4 0.419 0.051 0.12 7
SW (CP) 2 2.308 1.254 0.54 7
Environment for T-type joints with T=32mm under axial loading Air 3 0.166 0.060 0.36 18
SW (CP) 4 0.340 0.242 0.71 18

The above result can be explained by the high residual tensile stress expected in tubular welded joints, which made the stress ratio effect insignificant. Thus in the remaining assessments, R-ratio was not considered to be an influential factor on Re, and similar tests with different R-ratios were classified into the same group to increase the sample size for comparison.

Effect of tubular geometric parameters & chord thickness

A tubular joint geometry can be characterised by four parameters:

spyhzjune08e2.gif    

[2]

The following three geometric parameters have been analysed to assess their effects on remaining life of through-thickness cracked members: β with chord dimensions and τ constant; τ with chord dimensions and β constant; chord thickness with γ nearly constant.

Effect of β on Re

In the fatigue tests on 168mm and 457mm diameter T joints under out-of-plane bending (OPB) loading reported by Wylde[5,6], different brace diameters were used in the specimen design to investigate the effect of parameter β on fatigue endurance. For the smaller chord diameter with a chord thickness of 6mm, twelve specimens with two β values, 0.5 and 1.0, were tested.[5] Results of the assessments are given in Table 2. No marked influence of parameter β on Re can be seen under such testing conditions.

The above conclusion can also be seen in another size of T joint with D=457mm and T= 16mm, all specimens having a similar τ value[6], see Table 2. In this comparison, the mean Re value at β=0.25 was higher than that at β=1.0, but this is not considered to be a significant difference between the two cases.

From the above assessments, it appears that β is not a critical factor influencing the remaining life. In the following assessments therefore, no distinction among tests with different β values was made.

Effect of τ on Re

Eight T type joints with two different brace thicknesses in the specimen design were tested by NEL under axial loading.[12] All specimens had a chord diameter of 457mm and a chord thickness of 16mm. The assessment indicated that the specimens with a smaller brace thickness (τ = 0.55) had a higher mean Re value, see Table 2.

Under OPB loading, experimental data from tests[6] with the same joint type and chord dimensions as above suggested that the mean Re values of the specimens with a thinner brace wall (τ = 0.6) was significantly higher than that from specimens with a thicker brace wall (τ = 1.0), see Table 2.

The above comparisons suggest that the remaining life of through-thickness cracks decreases with increasing τ value and that this effect is more evident under OPB loading than under axial loading.

Effect of chord wall thickness on Re

In the following assessments, all specimens assessed had a chord radius to chord wall thickness ratio (parameter γ) of around 14.

T joints with two chord sizes, T=6mm and T=16mm, were tested by Wylde[5,6] under OPB loading. The attainment of a through-thickness crack in all investigations by Wylde[5-10] was found to be readily detected as a secondary maximum in strain range on the graph of strain output from the gauge element closest to the centre of the crack. For some tests, a small internal pressure of 5lbf/in was applied to the chord. Both methods gave very close results on N3 life. The end of the test was taken to correspond to the formation of a crack which had propagated approximately a distance equal to the chord radius away from the weld toe, at both ends of the crack.

The analysed results for the effect of chord size on Re are given in Table 2. It can be seen that chord thickness had a marked effect on Re under OPB loading, with the smaller tubular joints having a substantially higher mean Re value. Wylde[6] noticed that, unlike the results for the 6mm chord thickness joints which all failed in a similar fashion, the larger diameter joints (T=16mm) exhibited a wide range of cracking modes. These involved weld toe cracking on both the chord and brace side, and in many cases extensive inter-run cracking in the weld metal. Multiple crack initiation occurred in the weld ahead of the main crack and grew backwards to joint the main crack, which often branched away from the weld toe.

Discussion

The fatigue endurance of a welded member generally decreases with increasing thickness[25,26] and there is a thickness penalty on fatigue life in the current HSE fatigue design guidance[4] because of the effect of the high stress concentration and the low stress gradient across thickness section associated with larger tubular joints on crack initiation/early growth. For through-thickness crack growth, however, such effects do not exist at the whole crack front, thus the remaining life of through-thickness cracks was largely insensitive to chord thickness under axial loading for thickness up to 32mm.[13] Under OPB loading, there was not only a significant difference between the fatigue performance of the two sizes of joints, when compared at the same hot spot stress ranges, but there was also a substantial difference in Re values between the two sizes of joints. This can be clearly seen in Figure 4 where the mean N3 and N4 S-N curves of both sizes of joints[5,6] are compared.

Fig.4. Comparison of mean S-N curves for 168mm and 457mm diameter T joints tested under OPB loading [5,6]
Fig.4. Comparison of mean S-N curves for 168mm and 457mm diameter T joints tested under OPB loading [5,6]

The gap between N3 and N4 for specimens with T=16mm was much less than that for specimens with T=6mm. When compared at a similar hot spot stress (HSS) range under OPB loading, the crack growth rate of tubular joints with larger sizes (D=457mm and T=16mm) was found to be higher than that of smaller joints (D=168mm and T=6mm). This can be seen by comparing the crack growth curves of the joints at two sizes shown in Figure 5 for a HSS range of ~280MPa and in Figure 6 for a HSS range of ~310MPa. In Figure 5, all four specimens of two chord sizes had a similar crack length at N3, however the cracks in thetwo larger specimens grew faster than those in the smaller specimens. In Figure 6, the primary crack in specimen 704/1 with T=6mm[5] grew very slowly after through-thickness cracking because of multiple crack growth fronts along and away from the weld toe. Comparatively, the growth rate of specimen 22/1 with T=16mm[6] was faster than other specimens. It can also be seen from Figure 6 that there were variations in crack growth rate among these specimens even when compared at the same specimen size. This suggests that there are other factors, for example, crack growth path and crack coalescence, which can affect growth rate for through-thickness cracks.

Fig.5. Comparison of primary through-thickness crack growth against normalised cycles for two sizes of joints at hot spot stress range of ~280MPa [5,6]
Fig.5. Comparison of primary through-thickness crack growth against normalised cycles for two sizes of joints at hot spot stress range of ~280MPa [5,6]
Fig.6. Comparison of primary through-thickness crack growth against normalised cycles for two sizes of joints at hot spot stress range of ~310MPa [5,6]
Fig.6. Comparison of primary through-thickness crack growth against normalised cycles for two sizes of joints at hot spot stress range of ~310MPa [5,6]

Effect of loading mode

Comparison of axial and OPB loading modes

Data are available for assessment of axial[3,12] and OPB[5,6] loading modes on Re for T joints with two sizes of chord thicknesses: T=6mmand T=16mm. Since the remaining life of through-thickness cracks was chord thickness dependent under OPB loading as shown before, comparisons were hence made separately for each chord thickness, see Table 2. It can be seen that, with other conditions the same, OPB loading had a larger mean Re value than those under axial loading, especially at the smaller chord thickness where the mean Re value of the former was more than three times higher than that for the latter. This difference reduced with increasing chord thickness.

The longer remaining life of through-thickness cracks under OPB loading was also found in K and KT joints with a chord thickness of 16mm.[10] Details of cracking locations and crack development in these tests were reported.[10] The mean Re value under OPB loading was much higher than that under axial loading, see Table 2. As the sample size for the axial loading is small, the Re value obtained under this loading mode needs to be interpreted with caution.

Comparison of Axial and IPB Loading Modes

A series of fatigue tests on T joints with two sizes of chord thickness, T=6mm and T=16mm, were carried out by NEL[12] under both axial and IPB loadings, and some within the ECSC funded programme under axial loading.[3] The mean Re values under IPB loading mode were higher, especially at a chord thickness of T=16mm, see Table 2.

Discussion

Under axial loading, cracking occurred mainly along the weld toe, either on the chord side or on the brace side. Multiple crack initiation often occurred at these locations and frequently the growth of the primary crack involved coalescence with these small cracks resulting in comparatively fast crack growth and leading to final failure by the separation of brace from chord. A schematic illustrating this kind of crack development is shown in Figure 7,where the numbers in the crack growth paths indicate the sequence of crack length measurements. In this example, cracking occurred simultaneously in both chord and brace.[6] Under axial loading, crack growth rates after through-thickness cracking became faster as can be seen in the example in Figure 8 for a T joint with a chord thickness of 32mm.[7] This was inagreement with the general observation on the stiffness change of the specimens tested under this loading mode, where no change in specimen stiffness was observed until the final 10% of life. [9,10]

Fig.7. Crack development of an axially loaded specimen, showing cracking at weld toes on both chord and brace, crack growth branching and coalescence [10]
Fig.7. Crack development of an axially loaded specimen, showing cracking at weld toes on both chord and brace, crack growth branching and coalescence [10]
Fig.8. Primary through-thickness crack growth of 914mm diameter T joint under axial loading [7]
Fig.8. Primary through-thickness crack growth of 914mm diameter T joint under axial loading [7]

Under OPB loading, cracks generally initiated at the saddle position, at the weld toe on the chord side. They initially propagated along the weld toe and then branched into the chord, resulting in simultaneous growth into the chordand along the weld toe.[5,6]

A schematic illustration showing this kind of crack development is given in Figure 9. Under such a loading mode, the crack growth rates before and after through-thickness cracking did not change much. This can be seen from an example in Figure 10 by comparing the slopes of the crack length extension curve before and after through-thickness cracking. This reflects the gradual change in stiffness observed in tests under OPB loading.[9,10] It can also be seen from this figure that the primary crack propagated slightly faster when it just started branching away from the weld. However afterwards it had more or less the same growth rate as that before branching.

Fig.9. Crack development in a T joint with a chord diameter of 457mm under OPB loading [6]
Fig.9. Crack development in a T joint with a chord diameter of 457mm under OPB loading [6]
Fig.10. Primary through-thickness crack growth data of a 168mm diameter T joint under OPB loading [5]
Fig.10. Primary through-thickness crack growth data of a 168mm diameter T joint under OPB loading [5]

Compared at the same HSS range, the crack growth rate under axial loading was often found to be faster than that under OPB loading. This can be seen from Figure 11 which compares the crack growths of four specimens in K/KTjoints at a similar HSS range of ~190MPa, two under axial and two under OPB loading. It demonstrates that, under OPB loading, the crack growth rates were slower and the remaining lives of through-thickness cracks, as indicated by the ratio N/N3 at the end of the test, were longer. In one test under OPB loading, the primary crack branched away from the weld toe into the chord soon after it penetrated through the chord thickness. There was no secondary cracking involved during the crack propagation. The crack growth rate decreased as it grew far into the chord, see Figure 12, and even stopped growing when the actuator reached its limit.[6] The test was finally terminated. The development of this crack is schematically shown in Figure 13.

Fig.11. Comparison of primary through-thickness crack growth against normalised cycles for two loading modes at hot spot stress of ~190MPa [10]
Fig.11. Comparison of primary through-thickness crack growth against normalised cycles for two loading modes at hot spot stress of ~190MPa [10]
Fig.12. Primary through-thickness crack growth data of a 457mm diameter T joint under OPB[6] - crack branched away from weld into chord during its early crack propagation
Fig.12. Primary through-thickness crack growth data of a 457mm diameter T joint under OPB[6] - crack branched away from weld into chord during its early crack propagation
Fig.13. Schematic diagram showing the crack development of specimen 23/1 in[6] - crack growth of joint given in Figure 1
Fig.13. Schematic diagram showing the crack development of specimen 23/1 in[6] - crack growth of joint given in Figure 1

Crack shape obviously affects the Re value since it determines the through-thickness crack length at N3. It was known that a semi-elliptical crack could quickly develop into a straight-fronted crack once it penetrated through wall.[19,24] Crack shape development with cycles in tubular joints at several conditions has been reviewed by Tweed.[27] However, the effects of loading mode and small chord thicknesses (T<32mm) on crack shape were not reported. From the detailed records of the fatigue tests on K and KT joints of similar geometry under axial and OPB loadings by Wylde,[9,10] the mean ratio of crack depth to half crack length, a/c, at N3 of these specimens is summarised in Table 3. It can be seen that the cracks in the specimens tested under OPB loading tended to take a much shallower shape compared to those under axial loading. Although this suggests that OPB loading would produce a lower Re value, the mean Re value under OPB loading was in fact higher than that under axial loading due to the high crack propagation rate under axial loading as shown in Figure 11.

Table 3: Comparison of crack aspect ratio, a/c, at N3 for axial and OPB loading & for T, K and KT joints[9,10]

Chord dimensions
(mm)
Axial loadingOPB loadingT jointsK/KT joints
Sample sizea/c ratioSample sizea/c ratioSample sizea/c ratioSample sizea/c ratio
168 x 6 2 0.16 6 0.09 15 0.15 6 0.09
457 x 16 2 0.11 8 0.07 15 0.27 8 0.07

Effect of joint type

Although the majority of the tests in the database collected were for T joints, relevant data on K, KT, Y, X and Monopod types of joints are also available. Comparison of T joint with K and KT joints in the remaining life of through-thickness cracks is given below and the comparisons with other joints can be found in.[13]

Comparison of T and K/KT joints tested under OPB

A series of fatigue tests was conducted by Wylde[5-10] on T, K and KT joints under OPB loading at two chord thicknesses, 6mm and 16mm. To increase sample size for statistical reliability, tests from K and KT joints with and without brace overlapping were treated as one group since the majority of these joints had overlapped braces. The Re values for T and K/KT joints were significantly different when compared at a chord thickness of 6mm, with the mean Re value for T joints being about triple that for K and KT joints, see Table 2. Wylde[7] noticed that the N4 lives of T and K/KT joints were almost the same when they were compared at the same hot spot stress range. However, the average N3/N4 ratio was 0.492 for T joints while it was 0.741 for K and KT joints. When chord thickness was increased to 16mm, the mean Re value increased for K and KT type joints but decreased for T type joints, see Table 2. At such a chord thickness, the difference in mean Re values between T and K/KT joints was small.

Discussion

In the review of crack shape development in tubular joints, Tweed[27] found the cracks in K and Y joints under axial loading tended to take a shallower shape when compared to T joints. The average a/c ratio at N3 was about 0.09 for Y and K joints, while it was about 0.14 for T joints. Similar results were also found in the tests under OPB loading[5,6,9,10] for two sizes of tubular joints. Compared to T joints, the crack shape for K/KT joints was much shallower, see Table 3.

The reason for a low a/c ratio for K and KT joints under OPB loading can be explained by the fact that there were two peaks in the strain distribution around the weld toe, opposite the centreline of each brace. Cracks initiated at two hot spots and grew independently along the weld toe as a surface crack. Once they joined with each other, they immediately became a through thickness crack which had a length almost twice of that for equivalent T joints at N3.[10] It is interesting to note that, when plotted against HSS range, there was no significant difference between the fatigue endurance (N4) of T, K and KT joints under OPB loading.[10]

Effect of environment

K and KT Joints under OPB loading

Wylde[7] carried out fourteen fatigue tests on K and KT joints to investigate their fatigue endurance in air, in seawater with free corrosion (FC) and with cathodic protection (CP). All specimens had a chord diameter of 168mm and a chord wall thickness of 6mm and were tested under OPB loading. It can be seen from Table 2 that the mean Re value was the lowest for specimens tested in air while those specimens tested in seawater with CP had the highest remaining life.

T Joints under axial loading

Similar results were also found in tests by NEL on T joints with T=32mm.[18] All specimens were tested under axial loading. The mean Re value for specimens tested in seawater with CP was much higher than those specimens tested in air, see Table 2.

Discussion

In the review of crack shape development in tubular joints, Tweed[27] reported that a crack in a tubular joint tended to take a deeper shape (higher a/c) in seawater environment than in air, i.e., the crack length at N3 was shorter in seawater. This observation was also reported in another study of double T joints of 914mm chord diameter and 32mm chord wall thickness under IPB loading.[28] It was reported that the crack aspect ratios were significantly different in air and seawater. In seawater a single crack tended to dominate throughout the life and the a/c ratio was about 0.2 on breaking through the chord wall. On the other hand, in air, multiple cracking along the toe led to very low aspect ratios, about 0.15 at N3. The comparatively shorter crack length at N3 for specimens tested in seawater might be the reason for their higher mean Re value, although the crack propagation rates in seawater were generally faster than that in air. At the lower regime of stress intensity factor ranges, the crack propagation rates in seawater with CP and in air were almost identical.

Girth welds in plain brace members

There were very limited fatigue data containing N3 and N4 lives for girth welded joints in plain tubes. A recent study by TWI [24] investigated the remaining life of through-thickness cracks in girth welded pipes. All specimens, having an outer diameter of 324mm and a chord wall thickness of 12.7mm, were tested under four-point-bending loading. Through-thickness cracking was detected by an internal digital camera. The Re values are summarised in Table 4. It can be seen that the remaining lives were relatively short. It should also be noted that, in contrast to tubular intersections, cracking in girth welded pipes always started from the weld root and crack growth always followed the weld.

Table 4: Re for girth-welded pipes

Sample sizeMean ReSDCOVReference
7 0.086 0.04 0.46 24

General discussion

The development of cracks in tubular intersection joints under fatigue loading is complex and can be influenced by many factors. The remaining life of through-thickness cracks was found to depend on specimen geometry and testing conditions. However, it should be noted that there were some dissimilarities between the experimental data examined and the actual situation in a structure with respect to the remaining life of through-thickness cracks. The following three aspects should be noted:

  1. Reduced stiffness due to cracking results in changing load paths within the structure and possible load shedding away from the damaged area. Such shedding can reduce the stress range and stress intensity factor K, which theoretically could drop below the threshold value and the crack might stop growing. However, this will also result in enhanced stresses elsewhere and hence accelerate fatigue failure. These effects are not reflected in simple laboratory tests.
  2. Most of the tests with N3 and N4 lives were conducted in air, not representing those offshore members which are to be inspected by FMD. In submerged members, the remaining life of through-thickness cracks might be longer than those tests in air because of environmental effects on the crack shape, as discussed before. However, the data are very limited.
  3. The actual crack shape in members of offshore structures might be of lower aspect ratio than that observed in laboratory testing. Multiple loading paths in complex nodes and the combination of axial, in-plane and out-of-plane loading modes all contribute to a greater number of hot spot locations in a joint, resulting in multiple initiation and lower aspect ratios. The low aspect ratios correspond to significant reduction of the load-bearing area. It was shown in [1] that at Re = 0.2 the aspect ratio, a/2c, typically lies in the range 0.03 - 0.06 and that the corresponding cracked area is between 20% and 50% of the total uncracked area, depending on joint type. This results in significant reduction of the static capacity. Thus, consideration needs to be given to the loss of static strength, as well as fatigue strength, in a structural integrity management plan.

To enable the application of the knowledge generated from the current assessment, reliability analysis can be used in conjunction with fracture mechanics to evaluate the effect of FMD on structural integrity by the prediction of the remaining life beyond N3, the change in probability of failure with time and the subsequent determination of a suitable FMD inspection interval. The justification for the use of FMD can be made by comparison of the through-life structural reliability against a target structural reliability and ensuring that, for the relevant inspection interval that the target, or change in reliability over the period, is not exceeded. Such an analysis needs to take into consideration both the fatigue and ultimate limit states and can be rather complex.

Ultimate failure can be defined assuming either component failure (based on code checks) or system failure (based on non-linear pushover analysis). The fatigue assessment needs to quantify the behaviour of cracks prior to and after they penetrate the wall thickness to accurately evaluate the use of FMD. Fracture mechanics procedures, e.g. BS 7910[29], provide the means to do this, permitting the calculation of the necessary number of cycles at a given stress range for a deep surface crack to grow from a particular a/T and a/2c to the through-thickness condition and subsequently on to failure.

Until recently, stress intensity factor solutions were not available for very deep surface cracks approaching a/T=1 and plastic yield collapse load solutions for circumferential surface or partly through-thickness cracks in tubular members are subject to considerable uncertainty. However, both stress intensity factor and plastic collapse solutions were studied in a joint industry project which developed guidance on the use of FMD[30] for the application of fracture mechanics to the assessment of FMD. Stress intensity factor solutions were derived for very deep surface circumferential cracks, part through-thickness cracks and through-thickness cracks in tubes in tension.

Conclusions

In total, 281 test results with both N3 and N4 fatigue lives were collected. The data cover a variety of joint geometries, joint types, loading modes and many other testing conditions. Because of the complexity of the interactions of many parameters on remaining life of through-thickness cracks, as well as the limited number of tests in some conditions, it is difficult to draw sound conclusions for some cases. Nevertheless, compared to the reference condition, the following conclusions/trends are tentatively drawn based on the assessments:

  • taking all the results together, the mean value of Re (the remaining life of a through-thickness cracked member) was 0.443. The ratio N4/N3 is 1.38 and 1.25 for the mean and mean-2SD curves, respectively. However, the data were widely scattered (see Figures 1-3).
  • under axial loading stress ratio did not show a significant influence on Re.
  • β did not exhibit a significant effect on Re under axial and OPB loading.
  • Re was found to decrease with increasing values of τ, especially under OPB loading.
  • Re was found to be dependent on chord wall thickness, decreasing under OPB loading.
  • compared to axial loading, higher Re values were achieved under OPB and IPB loadings, especially under OPB loading for joints with a small chord size.
  • under OPB loading, T joints achieved a higher Re value than K/KT joints when compared at a chord wall thickness of 6mm. However the difference in Re between the two types of joints became small when compared at a chord thickness of 16mm.
  • specimens tested in free seawater corrosion and cathodic protection had a higher Re value.
  • girth welded joints in plain tubes gave a low Re value, with a mean of only 0.086.

The data show the wide variability of N4 fatigue lives relative to N3. Some trends have been identified and the data presented in this paper should enable better prediction of the fatigue performance of tubular joints, resulting in enhanced structural integrity management of offshore installations. However, the data also demonstrate that there is a need to better understand the factors that influence the behaviour of through-thickness cracks for the effective integrity management of offshore installations. This entails not only an understanding of the influence of the parameters discussed in this paper but also that of the system behaviour, both in the undamaged and the damaged state. This is a complex area and is currently being researched by HSE.

Acknowledgements

The views expressed in the paper are those of the authors. The authors would also like to thank Dr. Graham Wylde, TWI for help in supplying some of the data and Dr Geoffrey Booth, TWI for helpful discussions.

References

  1. Sharp J V, Stacey A and Wignall C M, 'Structural integrity management of offshore installations based on inspection for through-thickness cracking', 17th International Conference on Offshore Mechanics and Arctic Engineering, OMAE98-2110, 1998.
  2. Raine G A, 'The changing face of testing of oil and gas offshore installations', Materials Evaluation, 58, 2002, pp.951.
  3. Department of Energy, 'Background to new fatigue design guidance for steel welded joints in offshore structures', Report of the Department of Energy Guidance Notes Revision Draft Panel, 1984, HMSO.
  4. HSE, 'Guidance on design, construction and certification', 4th Edition, London U.K., January 1990 (including amendment No.3 published in Feb. 1995).
  5. Wylde J G, 'Fatigue tests on 168mm diameter tubular T-joints under out-of-plane bending', TWI Report 3612/1/82, February 1982 (produced under contract to UKOSRP).
  6. Wylde J G, 'Fatigue tests on 457mm diameter tubular T-joints under out-of-plane bending', TWI Report 3612/2/82, December 1982 (produced under contract to UKOSRP).
  7. Wylde J G, 'Fatigue tests on welded tubular joints in air and sea water', TWI Research Report 251/1984, 1984 (confidential to Industrial Members of TWI).
  8. Wylde J G, 'Seawater corrosion fatigue tests on steel tubular joints', Internal TWI Report 7921.01/87/589.2, 1988.
  9. Wylde J G, 'Fatigue tests on 168mm diameter welded tubular K and KT joints under out-of-plane and axial loading', TWI Report 3612/3/83, April 1983 (produced under contract to UKOSRP).
  10. Wylde J G, 'Fatigue tests on welded tubular K and KT joints of 457mm diameter out-of-plane and axial loading', TWI Research Report 3612/4/84, July 1984 (produced under contract to UKOSRP).
  11. Eide O I and Berge S, 'Fatigue capacity of stiffened tubular joints', 9th International Conference on Offshore Mechanics and Arctic Engineering, Vol.3, 1990, pp.209.
  12. Macdonald A, Brown C M, Thomson J F and Kerr J, 'Strain distribution measurement and fatigue tests on welded tubular joints', Report No. B12 in United Kingdom Offshore Steels Research Project - Phase I Research contract reports, Department of Energy, OTI 89 540,1988.
  13. Zhang Y H and Wintle J B, 'Review and assessment of fatigue data for offshore structural components containing through-thickness cracks', TWI report for HSE, April 2004.
  14. Back J de and Vaessen G H G, 'Effect of plate thickness, temperature and weld toe profile on the fatigue and corrosion fatigue behaviour of welded offshore structures', Final report, Part II, ECSC Convention 7210-KG/601, Delft/Apeldoorn May 1984.
  15. Damilano D F, Camisetti C and Negri A, 'Fatigue behaviour of unstiffened and stiffened Y tubular joints', Steel in Marine Structures, Paris, Paper No. 10.1, 1981.
  16. Maosheng G C, Shan W and Shaohu U L, 'Crack growth and fatigue life estimation in welded T-tubular joints based on fracture mechanics', 9th International Conference on Offshore Mechanics and Arctic Engineering, Vol.3, pp.351, 1990.
  17. Department of Energy, United Kingdom Offshore Steels Research Project - Phase II, Final Summary Report , OTH 87 265, 1987.
  18. Department of Energy, United Kingdom Offshore Steels Research Project - Phase II Progress Report No.12, UKOSRP II/PSC/P85-2T, from 16/9/84 to 5/1/85.
  19. Jiao O G, Skallerud B and Eide O I, 'Residual fatigue life of tubular joints with through thickness cracks', SINTEF Report STF71 F89063, Trondheim, Norway, 1990.
  20. Lourenssen A A and Dukstra O D, 'Fatigue tests on large post weld heat treated and as welded tubular T-joints', Offshore Technology Conference 4405, pp.327, Houston, May 1982.
  21. Maddox S J, Wylde J G and Yamamoto N, 'The significance of weld profile on the fatigue lives of tubular joints', 14th International Conference on Offshore Mechanics and Arctic Engineering, Vol.III, pp.127, ASME, 1995.
  22. Sanz G, Lieurade H P and Gerald J, 'Fatigue tests on ten full-scale tubular joints', Steel in Marine Structures, English version of French paper, paper 8.1, Paris, October 1981.
  23. Gowda S S, Arockiasamy M, Reddy D V and Muggeridge D B, 'Corrosion fatigue strength of offshore monopod tubular joints in cold ocean environment', Offshore Technology Conference 4523, pp.93, Houston, May 1983.
  24. Pereira M, 'Growth of through wall fatigue cracks in brace members', TWI Report 13467/3/02, HSE report RR 224, HSE Books, 2004.
  25. Sharp R P, 'Fatigue strength of transverse fillet welded joints in steel subjected to bending', Engineering Science Data Unit, June 1978.
  26. Gurney T R, 'The influence of thickness on the fatigue strength of welded joints', Second International Conference on Behaviour of Offshore Structures, London, August 1979.
  27. Tweed J H, 'Remaining life of defective tubular joints: depth of crack growth in UKOSRPII and implications', Department of Energy Offshore Technology Report OTH 87 278, HMSO.
  28. Tubby P J, Eide O I and Skallerud B, 'Variable amplitude fatigue of steel tubular joints in seawater with cathodic protection', 13th International Conference on Offshore Mechanics and Arctic Engineering, Vol.III, 1994, pp.181.
  29. BS 7910:2000, 'Guide on Methods for Assessing the Acceptability of Flaws in Fusion Welded Structures', British Standards Institution, London, 2000.
  30. EQE International Ltd., 'Guidance on the Use of Flooded Member Detection for Assuring the Integrity of Offshore Platform Substructures, HSE Report OTN 2000 041, 2000.

For more information please email:


contactus@twi.co.uk