YanHui Zhang, Simon Smith, Liwu Wei and Carol Johnston
TWI Ltd
Cambridge, United Kingdom
Alexander Stacey
Health & Safety Executive London, United Kingdom
Paper presented at Proceedings 32nd International Conference on Ocean, Offshore and Arctic Engineering (OMAE 2013) June 914, 2013, Nantes, France
Abstract
Residual stresses (RS) are unavoidably generated in components after welding. Tensile residual stresses in engineering structures generally have an adverse effect on structural integrity, with detrimental effects on brittle fracture, corrosion properties and fatigue performance.
Residual stress measurements for offshore welded joints are generally very limited. In this study, experimental measurements and numerical modelling of residual stresses were carried out and the throughthickness residual stress distributions in circumferential butt welds are presented. Pipes of two different dimensions and steels were investigated: a seamless X70 steel pipe with an outside diameter of 406mm and a seamwelded X65 steel with an outside diameter of 508mm. The measurement methods used included the centrehole air abrasion drilling, the deephole drilling (DHD) and the block removal, splitting and layering (BRSL). The numerical modelling was undertaken to determine whether modelling could provide a satisfactory prediction of the final residual stresses. Finally, the new experimental data were combined with the BS 7910 database to derive revised upper bound throughthickness residual stress distributions for girth welds and to provide an assessment of the impact of residual stresses on the structural integrity of circumferential welds in offshore structures.
Introduction
Residual stresses are unavoidably generated in components after welding. The magnitude of residual stresses may be up to the yield strength of the material. Tensile residual stress in engineering structures generally has an adverse effect on structural integrity. It can cause detrimental effects on brittle fracture, corrosion properties and fatigue performance.
Published experimental data from residual stress measurements are limited (14) and are mostly for stainless steels for the nuclear industry (5,6). The experimental data for offshore welded joints are very limited. There is strong interest in the offshore industry to better understand the magnitude and the distribution of throughwall thickness residual stresses in order to justify, improve and supplement the guidance on residual stresses described in codes such as BS 7910 (7). Increased knowledge of residual stresses can help the offshore industry gain more confidence in structural integrity of critical components.
A comprehensive study of residual stresses in offshore girth welded joints, which was funded by the health and safety executive (HSE) was carried out recently. It included a brief literature review, surface and throughwall thickness residual stress measurements in 16inch and 20inch diameter pipes, numerical modelling of residual stress profiles, determination of an upper bound expression for the throughthickness residual stress distribution for girth welds based on the revised database and assessment of the impact of the results on structural integrity assessment. This paper describes the results obtained and the revised upper bound throughthickness residual stress distributions proposed for girth welds in offshore structures.
Residual Stress Measurements
Residual Stress Measurements for the 16 inch Pipe
The material was 406mm (16 inch) outside diameter (OD), 19.1mm wall thickness (WT) and 0.9m long steel pipe to API 5LX70 specification. The yield and tensile strengths of both the base and weld metals were determined experimentally and the results are given in Table 1. All welds were made by Heerema Marine Contractors (HMC) using a typical steel catenary riser (SCR) specification, in 5G position. The weld profile is shown in Figure 1(a). Details of the fabrication conditions are summarised in Table 2.
TABLE 1 Tensile properties of the base and weld metals of the two girth welded pipes
Pipe diameter, mm

Metal

Yield strength MPa

Tensile strength MPa

Elongation %

Reduction area, %

406 (16 inch)

Base

519.0

601.0

19.4


Weld

728.0

782.1

18.1

64.5

508 (20 inch)

Base

539.2

615.9

18.9


Weld

653.7

720.5

19.0

69.8

Note: Base metal perpendicular to weld; Weld metal parallel to weld.
TABLE 2 Details of fabrication for the 16 and 20 inch girth welded pipes
Pipe size 
Pass

Process

Travel speed, cm/min

Arc energy, kJ/mm

406mm (16 inch) OD x 19.1mm WT in 0.9m length, seamless

1

PGMAW

80

0.37

2

PGMAW

46

0.62

3

PGMAW

46

0.61

4

PGMAW

46

0.64

5

PGMAW

45

0.64

6

GMAW

36

0.69

7

GMAW

30

0.71

508mm (20 inch) OD x 22.9mm WT in 0.9m length, seam welded UOE

1

PGMAW

79

0.33

2

PGMAW

55

0.54

3

PGMAW

53

0.55

4

PGMAW

51

0.58

5

PGMAW

49

0.60

6

PGMAW

47

0.61

7

GMAW

40

0.62

8

GMAW

31

0.66

The centrehole air abrasion drilling method (8) was used for residual stress measurements near surfaces, with the holes located at either the weld cap toe or the weld root toe, which are typical fatigue crack initiation sites. a total of eight measurements were performed on three ring samples (ring ID: 1, 5 and 6), each being 0.9m long. for ring 1, four measurements were carried out: weld cap and weld root at both weld start and stop positions. For rings 5 and 6, the measurements were carried out at a position 60o from the weld start position, Figure 2. The results are presented in Table 3. The following can be seen from the table:
Residual stresses in the hoop direction (parallel to the girth weld) were very high, the largest being about 90% and 64% of the yield strengths of the base and weld metals, respectively. The levels of the residual stresses at the weld start and stop positions were higher than those at other positions. At both the weld start and stop positions, the residual stress was markedly higher at the weld root than at the weld cap.
FIGURE 1 Macrosections of the girth welds in the 16 inch and 20 inch pipes: a) 16 inch
FIGURE 1 Macrosections of the girth welds in the 16 inch and 20 inch pipes: b) 20 inch
Figure 2 Schematic, showing the locations for the residual stress measurements: 16in pipe on the left and the 20 inch pipe is on the right
 Residual stresses in the pipe axial direction were tensile at the weld root but compressive at the weld cap at the weld start and stop positions. However, the opposite was seen at the position 60° from the weld start. As this position represents the majority of the weld around weld circumference and the axial residual stresses at the weld cap were higher than those at the weld root, it is not surprising to see the angular misalignments towards OD side observed during extraction of the strip specimens due to residual stress relaxation (9).
 The results obtained from Ring 5 agreed well with those obtained from Ring 6 in terms of stress level and the sign of residual stresses on both weld cap and weld root sides. This suggests that (i) at locations away from the weld start/stop positions, there was less scatter in surface residual stresses between different welds; (ii) the centrehole method provided consistent results of residual stress measurements for similar locations.
TABLE 3 Results of the surface residual stress measurements for the 16inch girthwelded pipe
Sample

Location

Residual stress levels, MPa, and directions

Max principal stress

Angle to the axial direction

Axial stress

Hoop stress

Ring 1

Weld cap toe, 12 o’clock

290

74

21

263

Weld root toe, 12 o’clock

360

90

54

360

Weld cap toe, 6 o’clock

314

66

126

227

Weld root toe, 6 o’clock

469

89

111

469

Ring 5

Weld cap toe, 2 or 10 o’clock

244

67

114

222

Weld root toe, 2 or 10 o’clock

195

90

109

195

Ring 6

Weld cap toe, 2 or 10 o’clock

214

63

104

186

Weld root toe, or 10 o’clock

147

87

137

146

Notes: The 12 and 6 o’clock positions corresponded to the weld start and stop respectively; 2 or 10 o’clock positions cannot be uniquely identified  it depends on the side on which the weld is viewed.
Throughwall thickness residual stress measurements were carried out on one ring sample (No. 5) using the deephole drilling (DHD) method (10). Measurements were made at three locations: weld start, weld stop and 3 or 9 o’clock (Figure 2), the definition of this position depending on which side the weld is viewed. The measurement results are shown in Figure 3(a) for the axial stresses and Figure 3(b) for the hoop stresses. The axial and hoop stresses were normalised respectively by the yield strengths of the base and weld metals as they are relevant to the assumption about residual stresses in BS 7910, and the distance from the inside surface was normalised by the wall thickness of each pipe. The residual stresses at the surfaces measured by the centrehole method were also included for comparison.
FIGURE 3 Throughthickness distributions of the residual stresses and the residual stresses measured near the surface for the 16 inch pipe: a) Axial stress, parallel to the pipe
FIGURE 3 Throughthickness distributions of the residual stresses and the residual stresses measured near the surface for the 16 inch pipe: b) Hoop stress, along circumference of pipe
In general, the following features can be seen from the two figures:
 With respect to the residual stresses in the axial direction of the pipe, the stresses varied with position near the weld cap, with the stress being the greatest at the weld start position. The residual stress in all three positions increased and then decreased sharply with increasing distance from the inside surface of the pipe. It reached the maximum tensile residual stress at a depth of around 14mm. Near the weld root, all three locations showed compressive stresses and the magnitudes of the residual stresses at the three positions were similar. Except for the 3 o’clock position near the weld root location, correlation between the residual stresses obtained using deephole and centrehole drilling methods for other positions (weld start, stop and 3 o’clock) and locations (weld cap and root) was poor. The average axial residual stress over half of the wall thickness on the weld cap side was much higher than that on the weld root side, which was in agreement with the angular misalignment seen during extraction of the strip specimens from pipes.
 With respect to the residual stresses in the circumferential direction of the pipe, the stresses also varied with different positions near the weld cap, with the stress being the greatest at the weld start position, similar to that for the longitudinal stress. The residual stress in all three positions increased and then decreased gradually with increasing distance from the inside surface of the pipe. It reached the maximum tensile residual stress at a depth of around 13mm. The residual stresses in the circumferential direction were tensile throughout the wall thickness in all three positions with membrane stress predominant. Better agreement in residual stresses was seen between the deephole and centrehole methods for the circumferential stress. However, there was still lack of a good correlation for weld start position at the weld cap and 3 o’clock position at the weld root.
Residual Stress Measurements for the 20 inch Pipe
The material used was 508mm (20 inch) OD by 22mm WT and 0.9m long steel pipe to API 5LX65 specification. The pipe was produced by the UOE process and therefore it contained a longitudinal seam weld. The yield and tensile strengths of both the base and weld metals were determined experimentally and the results are given in Table 1.
All welds were made by HMC using the typical welding procedure for SCR. The weld profile is shown in Figure 1(b). Details of the fabrication conditions are summarised in Table 2.
Residual stresses at the surface of the pipe were measured at several locations and the centrehole technique was again used. Two positions were measured: one near the intersection of the girth and the seam welds and the other directly opposite to the first position (away from the seam weld). At each position, measurements on both inside and outside surfaces were conducted.
TABLE 4 Results of surface residual stress measurements for the 20 inch pipe, 2mm from weld toe
Location

OD/ID

Residual stresses measured using centrehole, MPa, and directions

Estimated from throughwall distribution

Max principal stress

Angle α^{2}, ^{o}

Axial stress

Hoop stress

Axial stress

Hoop stress

Near S/G intersection^{1}

OD

263

105

107

252

420

469

ID

451

93

182

450

349

200

Away from S/G intersection^{1}

OD

246

91

18

246

445

454

ID

288

92

54

287

391

272

Notes: 1. S/G: seam/girth weld; 2. α is the angle between the maximum principal stress direction and the axial direction of pipe.
Table 4 presents the residual stress measurement results obtained on the surfaces of the pipe sample. The maximum principal stress, the angle of the maximum principal stress to the pipe axis, the axial and hoop stresses are all presented in the table. It can be seen that the tensile hoop stresses ranged from 246 to 450MPa with the maximum value (450MPa) being found at the weld root close to the intersection between the seam and girth welds. This maximum value was less than the yield strengths of both the base and weld metals, especially for the latter (654MPa) which is more relevant for hoop stress (7). The axial stresses were lower than the hoop stresses, with a range of 54 to 182MPa.
FIGURE 4 Throughthickness distributions of the residual stresses and the residual stresses measured near the surface for the 20 inch pipe: a) Axial stress, parallel to the pipe
FIGURE 4 Throughthickness distributions of the residual stresses and the residual stresses measured near the surface for the 20 inch pipe: b) Hoop stress, along circumference of pipe
The throughwall distribution of residual stresses at the girth weld in the pipe sample was measured using the block removal, splitting and layering (BRSL) method. The method was pioneered by Rosenthal and Norton (11) in the 1940s, and has been further developed and in regular use at TWI since 1978. As for the surface measurements, the measurements of throughwall residual stress distributions were also made in two positions: the first was near the intersection between the seam and girth welds while the second was almost opposite to the first one. The measured throughwall residual stress distributions at the two positions are presented in Figure 4: Figure 4(a) for the axial stresses and Figure 4(b) for the hoop stresses. As for the 16 inch pipe, the axial and hoop stresses were respectively normalised by the yield strengths of the base and weld metals of the 20 inch pipe. It can be seen from Figure 4(a) that the patterns of the axial residual stress distributions from the two positions are clearly similar, both showing that compressive residual stresses existed over the central portion of the wall (between approximately 5 and 15mm from the bore of the pipe) while tensile stresses were observed over the portions near the inner and outer surfaces. Furthermore, the magnitudes of the residual stresses at the two positions were comparable with both being lying approximately between a compressive stress of 220MPa and a tensile stress of 600MPa.
Although the hoop residual stresses exhibited an irregular pattern with increasing distance from the inside surface, they were predominantly membrane tensile stresses. The hoop residual stress ranged from 151 to 710MPa. Similar features were observed between the two positions measured, despite some differences in magnitude.
Comparing the distribution patterns between the axial and hoop stresses, the hoop stress distributions obtained in the two positions can be largely characterised as “bend type” whose net effect is the production of a bending moment plus a membrane stress, while the axial stress distributions can be characterised more as ‘selfequilibrating type’.
The residual stresses at surfaces were estimated from the throughwall distributions and presented in Table 4 for comparison with those determined by the centrehole drilling method. A comparison is also made by including the surface residual stresses in Figure 4. It can be seen from both the table and figure that, although the surface stresses from the two methods were generally in agreement in indicating the nature of the stress (ie tension or compression), some large differences in magnitude were observed. For example, in the area near the intersection between the seam and girth welds, the hoop stress and the axial stress at the outer surface, as determined with the centrehole method, were 252 and 107MPa respectively, while the counterparts estimated from the throughwall distribution were 469 and 420MPa, respectively. It should be noted that the differences were not only due to the two different measurement methods, but also due to the different locations of measurement (there were approximately 15mm between the locations for the centrehole and BRSL).
Numerical Modelling of Residual Stresses in girth welds
Material Properties
Heat losses by convection and radiation from all external surfaces were modelled. The thermal properties were assumed to be temperature dependent with room temperature values as follows: specific heat capacity 450J/kgK; thermal conductivity 63W/mK. The pipe was assumed to have a density of 7800kg/m^{3}.
The stress–strain curves have been experimentally determined for both the base and weld metals for each type of pipe. These curves were used to provide the room temperature material behaviour. Data from the literature was used to estimate the dependence of the yield and ultimate strengths upon temperature. Linear kinematic hardening was assumed between the yield and UTS values. It was assumed that the UTS was reached at a strain of 10%. The temperature dependence of the Young’s modulus was taken from measurements (12). The room temperature thermal expansion coefficient was assumed to be 12x106/K and the value was also assumed to be temperature dependent.
It can be seen from the macrosections of the girth welds in Figure 1, the 20 inch girth weld was completed in eight passes while the 16 inch girth weld was completed in seven passes. Each pass was completed in two separate welds. The first weld laid down the pass on one side of the pipe from the start to the end 180° away. The second weld started and ended at the same location on the pipe, traveling in the opposite direction.
Modelling for the Two Pipes
The modelling was done using the commercial FE software Abaqus supplied by Simulia. Abaqus has heat transfer and nonlinear thermal stress analysis facilities which were used to make predictions of the welding heat flow and of the thermal stresses around the weld and therefore the resulting welding residual stresses. A model of the pipe with all of the weld metal was made. All of the elements representing the weld were removed at the start of the analysis and then sequentially restored at the point when the weld metal would have been deposited in the weld. This included the motion of each of the weld passes around the weld preparation circumference, as well as the progressive increase of the depth of the weld for each pass in sequence. The method meant that both the thermal capacity and stiffness of the material in the region of the weld was correctly built up during the welding process.
The heat flow model was solved in a sequence of steps and each step was split into smaller increments. An element was added at the start of a step to represent the new material added during the period of the step. A volumetric heat flux was added uniformly into the element during the whole period of the step such that the energy deposited was equal to the energy deposited in the same length of material in the actual weld. This approach achieved a correspondence between the model and the actual welding process. The time stepping continued in one direction for the first side of each weld from the start (12 o’clock) location to the stop (6 o’clock) position. The second side was then made using the same procedure from 12 to 6 o’clock, but welding in the other direction.
A continuous thermal stress analysis was run for the complete cycle of heating from the start of welding until the end of the last bead when the pipe cooled to ambient. The same process of adding elements individually through the same sequence and at an identical timing was done. This meant that each added element was immediately heated to the melting temperature or above because it represented the newly deposited molten material from the electrode.
A symmetric half of the full weld was analysed. The plane of symmetry was through the centre of the complete weld and perpendicular to the axis of the pipe. The displacements normal to the plane of symmetry of all nodes on the plane of symmetry were constrained. Displacements in the vertical direction (the direction from 6 to 12 o’clock) were constrained to prevent linear and rotational rigid body motion. A further constraint in the direction from 3 to 9 o’clock was also imposed to prevent rigid body motion in this direction. The full set of constraints therefore allowed the representation of the full pipe to deform freely under the action of the welding thermal stresses.
Results for the 16 inch Pipe
The predicted axial and hoop stress distributions at three positions are shown in Figure 5. As before, both the stresses and distance from the inside surface were normalised. The plots show that the axial stresses at the start and stop locations are similar, but they are different from the distribution at 3 o’clock. The through thickness distribution at 3 o’clock is an equilibrium distribution (the area of tension equals the area of compression). This must occur because over large regions of the weld the axial stresses in the whole weld must have an area sum of zero. Small regions can show nonequilibrium distributions, as revealed at the start and stop locations where the stresses are more compressive than tensile.
The throughthickness distributions of the hoop stresses are more similar at the three locations. The predicted stresses at the weld root are more tensile at the 3 o’clock position than at the start and stop locations.
Figure 5 also includes the measurement results for comparison with the predictions. It is apparent that there are differences between the two sets, especially for the axial stresses nearer the root and the hoop stresses nearer the cap. However, it is interesting to observe that the measured axial stresses at this position do not represent an equilibrium distribution because there is more tension than compression. This means that the measured values cannot be representative of large proportions of the rest of the pipe. This could either be because the start and stop locations have a long range of influence or because the pipes being joined did not have both material and geometric axisymmetry. The model suggests that the start and stop regions of interest are relatively limited, so it is suggested that the nonequilibrium axial stresses at 3 o’ clock arise from nonaxisymmetric pipes. The model was geometrically axisymmetric so the differences between the predictions and measurements are likely to arise from a surmised difference in the prewelded shape of the pipes being joined. The distributions are, however, qualitatively similar. Both the axial and hoop stress distributions increase underneath the cap, reach a peak and then fall towards the root. The predictions, however, show an increase in stress at the root which is not seen in the measured distribution.
FIGURE 5 Comparison of the FEA prediction of throughthickness distributions of axial and hoop residual stresses at three locations, and the experimentally measured residual stresses, for the seven pass girth butt weld in a 16 inch pipe: a) Axial stress
FIGURE 5 Comparison of the FEA prediction of throughthickness distributions of axial and hoop residual stresses at three locations, and the experimentally measured residual stresses, for the seven pass girth butt weld in a 16 inch pipe: a) Axial stress; b) Hoop stress
The predicted stresses exactly at the 12 o’clock position are qualitatively comparable to the measured distributions, but there are some numerical differences, especially in the axial stresses. However, the distribution 12mm away from the exact start has been found to be in better agreement with the measurements. In practice, the start location will not be as exact as the position of the modelling start location. It is therefore entirely possible that the good agreement at the 12mm position represents a real correspondence between the model and the test result. A similar effect is noticeable at the stop location.
Generally, there is good agreement between the measured hoop stresses and the predictions between 5 and 15mm below the surface of the weld. The agreement of the axial stresses in this region is fair, though it is poor at the 3 o’clock position. This is probably due to the shape of the test pipe as already discussed. The predictions and DHD values do not agree so well at the weld cap and weld root. Generally, the weld caps were machined prior to application of DHD, so there may have been both a redistribution of stresses because of the material removed and the generation of new stresses because of the machining process. The differences at the weld root are consistent. The modelling tends to show increased stresses at the root and no such trend was measured. It is unclear why this difference exists.
FIGURE 6 Comparison of the FEA predicted and DHD measured throughthickness distributions of axial and hoop residual stresses for the seven pass girth butt weld in the 20 inch pipe: b) Hoop stress
Results for the 20 inch Pipe
The predicted residual stress distributions for the girth weld in the 20 inch pipe at the three positions are shown in Figure 6. The figure also shows the results of the measurements made by BRSL.
Generally, there is good agreement between the measured and predicted axial residual stresses between 5 and 18mm from the weld root. However, it is interesting to observe that the measured axial stresses do not represent an equilibrium distribution because there is more tension than compression. This means that the measured values cannot be representative of large proportions of the rest of the pipe as explained in the above section for the 16 inch pipe. The agreement between the measured and predicted hoop residual stresses was not good throughout the plate thickness. However, it has been found that the predicted hoop residual stress distribution only 6mm from the weld centre line is dramatically lower than the results on the weld centre line. This suggests that the difference between the measurements and predictions may be because the measured residual stresses are averaged over the width of a block which is 12mm (4). There is a much smaller difference associated with the axial residual stresses at these locations.
Derivation of Axial Residual Stress Distribution for Low Heat Input
Introduction
The new data generated in this study and collected from the literature were then used to:
 Compare with the guidance given in BS 7910.
 Determine the upper bound throughwall thickness residual stress distribution.
 Establish the upper confidence limits.
As the experimental data generated in this study were from welds made with low heat input and residual stresses in the axial direction are often more important in structural integrity assessment, the analysis has focused on residual stresses in this direction and on welds with low heat input. The database considered in the statistical analyses includes:
 The database used for BS 7910 low heat input. The axial residual stress profiles given in BS 7910 were developed at TWI as a result of a review of publications in the open literature and TWI’s own database. It should be noted that the proposed equations were obtained by fitting the upper bound curve to published data and not based on statistical analysis.
 The residual stresses measured in the current project for the 16 and 20 inch pipes. They included three through wallthickness distributions for the 16 inch pipe and two through wallthickness distributions for the 20 inch pipe.
 New published data for girth welds made with or close to low heat input (12,13).
Statistical Analyses
The residual stresses were normalised with the yield strengths of the relevant base materials and the distances from the inside surface were normalised with the relevant wall thicknesses.
FIGURE 7 Experimental data for girth welds. The residual stress distribution given in BS 7910 for low heat input is included for comparison.
Figure 7 shows all the experimental data considered in the analysis. The database is composed of a total of 296 data points. The BS 7910 curve with low heat input is also included for comparison. It will be seen that the BS 7910 curve generally represents an upper bound for these data. However, near the surface regimes (x/t between approximately 0.1 to 0.2 and 0.8 to 0.9), some data points are above the BS 7910 curve.
FIGURE 8 Comparison of the residual stress distributions between the experimental data with different confidence levels and the FE predictions for the 16 inch pipe.
The whole data set was statistically analysed to determine the mean and the upper confidence limits for 80, 85, 90 and 95%. Results of the statistical analyses, together with the FE predictions and the lowheat input curve from BS7910, are shown in Figure 8 for the 16 inch pipe and Figure 9 for the 20 inch pipe. It will be seen that the BS 7910 curve represents an approximately 80% upper confidence limit which is less than the 90% upper confidence limit found by Serco (14) when they analysed the database for BS 7910. It will be seen that the mean curve derived agrees reasonably well with the curves predicted by the FEA. The curve corresponding to the 80% upper confidence limit provides the upper bound for the FE predictions for the two pipes, with a more conservative margin for the 20in pipe.
FIGURE 9 Comparison of the residual stress distributions between the experimental data with different confidence levels and the FE predictions for the 20 inch pipe.
The results shown in Figures 8 and 9 suggest that the BS 7910 curve can still provide a good estimate for the upper bound residual stress distribution for girth welds with low heat input. However, the BS 7910 curve does not always agree with the distributions of the data. For example, the current BS 7910 profile underestimates the residual stresses near x/t=0.15 but overestimates the residual stresses in the midwall thickness regime. Furthermore, it would be better to establish the curve based on a 90% confidence limit. Therefore, an attempt was made to establish a new expression to replace the current curve given in BS 7910. The expression was based on the 90% upper confidence limit derived above. However, it was slightly modified by limiting the ratio of residual stresses to yield strength to a maximum value of 1.0. This applied to the few data points near x/t>0.7 where the values of the ratios were slightly greater than 1.0. This was done to ensure consistency with the guidance given in codes on residual stresses for level one assessment where residual stress is assumed to be of the magnitude of the yield stress (the simplest and most conservative approach).
The new expression for the upper bound residual stress distribution was thus obtained as:
Where S_{R} is the residual stress and S_{Y} is yield strength of the base metal. Since the mean curve was below the curve with 90% upper confidence limit by a factor of 0.72 (Figure 8), the mean curve proposed was obtained by shifting the upper bound curve, given in Eq.1, by a factor of 0.72, ie:
FIGURE 10 Comparison of the recommended upper bound (UB) residual stress (RS) distribution curve (red, solid line) with the database and the others.
Figure 10 shows the new expressions proposed, in comparison with the BS 7910 low heat input and the 90% upper confidence limit curves. It can be seen from the figure that the proposed upper bound curve.
 Represents the combined data better than the current BS 7910 curve in that: (i) it increases the residual stresses at x/t=0.15; (ii), it reduces the residual stresses in the midwall thickness regime.
 Represents an approximately 90% upper confidence limit for the whole data (except for in the x/t=0.8 regime).
FIGURE 11 Throughwall thickness distributions of normalized stress intensity factor (SIF) K under axial residual stresses.
Stress Intensity Factor due to Residual Stresses
To investigate the effect of residual stresses on crack growth and fracture, stress intensity factor, K, was calculated for a series of crack sizes assumed. The calculations were carried out using the weight functions (15) which was incorporated in the TWI software Crackwise 4 with a special version. Each residual stress distribution was fitted to a polynomial equation of order 4 and the primary stress was assumed to be zero. A semielliptical surfacebreaking crack was assumed and the K value for each crack size was calculated at the deepest point. Figure 11 shows the normalised K values (normalised by σ_{y}√Πa) as a function of normalised crack depth for five different residual stress distributions: the mean and upper bound curves proposed, the 90% upper confidence limit determined from the regression analysis, the upper bound curve given in BS 7910 Annex Q and a uniform residual stress distribution with a magnitude of yield strength of material – the simplified approach given in BS 7910. In all calculations, a shallow crack aspect ratio, a/2c, of 0.1 was assumed and the crack was assumed to be at the weld root. It can be seen that the:
 K values corresponding to the proposed mean RS curve were low and decreased gradually with increasing crack size and approaches zero;
 Curves corresponding to the four upper bound RS distributions are significantly greater than that corresponding to the mean RS curve proposed;
 Curve corresponding to the proposed upper bound RS distribution is slightly lower than that corresponding to the 90% upper confidence limit;
 Curve corresponding to the proposed upper bound RS is higher than that corresponding to BS 7910 Annex Q when a/t<0.5, but slightly lower after a/t>0.5;
Normalised K values corresponding to the uniform RS distribution are considerably greater than those of other curves for all crack sizes. The overconservatism with this simplified approach given in BS 7910 can be appreciated by the differences from other curves.
Discussion
Comprehensive residual stress measurements were carried out in this project. For the 16 inch pipe, the results from the centrehole measurements suggested that the residual stress in the axial direction depended on the position along the circumference of a weld. It was higher (more tensile) at the weld root than at the weld cap at the start/stop positions. High tensile residual stresses on the surface of the weld root were also predicted by the FEA both in the present project and in other research (4). As a result, the effective mean stress level (effective mean stress = mean stress applied + residual stress) and hence the stress ratios would be higher at the weld root than at the weld cap in fatigue tests. Fatigue tests had been carried out on these girth welds and fatigue cracking had almost always initiated from either the weld start or the stop positions (16). As both crack growth threshold and crack growth rates strongly depend on the effective stress ratio, which was higher at the weld start/stop positions, it is possible that residual stresses were part of the explanation for the fact that the weld start/stop positions were the preferred sites for fatigue crack initiation from the weld root.
However, there was no clear effect of measurement location on residual stress for the seam‑welded 20 inch pipe. The RS results measured near the seam weld and at a position opposite to it were similar. This might be because of the processes following the seam welding, such as mechanical expanding (to achieve circularity), which may have diminished the difference in residual stresses at different positions.
It was noted that there was a disagreement between the centrehole and deephole measurements of axial residual stresses near the pipe surfaces for the 16 inch pipe. It is believed that the centrehole method would be more reliable for measuring nearsurface residual stresses (10). Indeed, VEQTER (17) agreed that the nearsurface results obtained from the deephole method should be taken with caution due to plastic deformation. Thus, the centrehole results should be considered to represent the residual stress state near the surfaces. With respect to residual stress distribution through the wall thickness, the results from the deephole method should be acceptable.
Residual stresses generally exhibit large scatter as can be seen from both the database for BS 7910 and the new database used in the current statistical analysis. This is because of many factors (3):
 Pipe wall thickness.
 Heat input per unit weld length (welding voltage, current and electrode speed).
 External restraints.
 Interpass temperature.
 Ratio of radius to thickness.
 Weld preparation (narrow gap, single or double V).
 Number of passes.
 Ambient environment (affecting cooling rate).
 Preheat.
 Phase transformation.
 Melting and remelting.
 Nonsymmetry (eg weld startstop, pipe roundness, pipe thickness mismatch, seam weld).
 Preexisting residual stresses due to steel making and manufacturing.
In the current BS 7910 guidance for axial residual stresses in girth welds, residual stress distributions are considered to be a function of heat input: low heat input (≤50J/mm^{2}), medium heat input (between 50120J/mm^{2}) and high heat input (>120J/mm^{2}).
Differences in residual stresses between measurements and numerical predictions have been found near the surfaces of pipes. It is suggested that this is probably due to the shape of the test pipes or the grinding of the weld cap (for the 16 inch pipe) which would result in a change of the original residual stresses. It should be noted that the residual stress levels at the start/finish were also predicted to be higher by FEA (4): tensile at the weld root but compressive at the weld cap.
The database for girth welds with low heat input has been updated. The statistical analysis indicated that the current residual stress profile given in BS 7910 for low heat input represents an approximate 80% upper bound confidence limit. Therefore, the new data will not have a significant impact on the current guidance on residual stresses and structural integrity assessment for girth welds. However, a better version of the throughwall thickness residual stress distribution was proposed which was based on the updated, larger database and an almost 90% upper confidence limit. Figure 11 suggests that, for surface cracks less than half of the wall thickness, it would be conservative to use the RS distribution proposed than the current BS 7910 curve. The simplified approach by assuming a uniform RS of yield strength magnitude in BS 7910 is unduly conservative.
Conclusions
For the 16 inch pipe
 There was disagreement in magnitudes of residual stresses measured between the centrehole and DHD techniques.
 The residual stress in the axial direction depends on position around the circumference of the weld. It was higher (more tensile) at the weld root than at the weld cap at the start/stop positions. However, away from these two positions the residual stresses were tensile at the weld cap but compressive at the weld root. The residual stress in the hoop direction of the pipe was tensile and predominantly a membrane stress
For the 20 inch pipe
 There was disagreement in the magnitudes of residual stresses measured between the centrehole and BRSL techniques.
 Similar distributions of throughwall hoop and axial stresses at the two locations measured were observed, implying a small influence of the seam weld on residual stresses in this case.
 The residual stress distributions in the axial direction exhibited a “U” shape: tensile near the inside and outside surfaces while compressive in the midwall thickness. It was different from that seen in the 16 inch pipe.
Numerical modelling
 For the 16 inch pipe, generally there is a good agreement between the measured hoop stresses and the predictions about 5mm below the surfaces of the weld. The agreement of the axial stresses in this region is fair, though it is poor at the 3 o’clock position. The modelling tends to show increased stresses at the root. This was in agreement with the surface measurement results (centrehole method) at the weld start and stop positions, but not with those measured by the DHD technique or the centrehole method at the 3 o’clock position.
 For the 20 inch pipe, generally there is good agreement between the measured axial stresses and the predictions about 5mm below the surfaces of the weld. Elsewhere there is a difference between the predicted and measured residual stresses. The predicted hoop stresses on the plane of symmetry of the model are higher than the measured values at this location. There appears to be a reasonably good correlation between the measured and predicted hoop residual stresses once the predicted distributions over the full width of the measurement block have been averaged.
Derivation of axial residual stress distribution
Based on the updated database, which included the database for the current BS 7910, the data experimentally obtained in the present project and the data collected in the literature, a new expression for residual stress distribution in the axial direction has been derived for welds with low heat input. It was based on a 90% upper bound limit (slightly modified).
Acknowledgement
The project was funded by Health and Safety Executive (HSE), UK. Its contents, including any opinions and/or conclusions expressed, are those of the authors alone and do not necessarily reflect HSE policy.
References
 Leggatt R H, 1982: ‘Residual stresses at circumferential welds in pipes’. The Welding Research Institute Bulletin, Vol. 23, pp181188.
 Leggatt R H, 1983: ‘Residual stresses at girth welds in pipes’. International conference on welding in energy related projects, Toronto, Canada, September
 Xu W, Leggatt R H, 2002: ‘Review of residual stresses in pipe girth welds’. TWI Single Client Project Report, TWI ref: 14127/1/02.
 Wei L, He W, 2010: ‘Comparison of measured and calculated residual stresses in steel girth and butt welds’. TWI Core Research Project report 942/2010 (confidential to TWI industrial members).
 Bouchard P J, 2007: ‘Validated residual stress profiles for fracture assessments of stainless steel pipe girth welds’, International Journal of Pressure Vessels and Piping, Vol 84, No. 4 pp 195222.
 Bouchard P J, 2001: ‘Validated residual stress profiles for fracture assessments of austenitic stainless steel pipe girth welds’. British energy report no. E/REP/ATEC/0025/GEN/01, December.
 BS 7910, 2005: ‘Guide on methods for assessing the acceptability of flaws in metallic structures’, British Standards Institution, London.
 ASTME83708, 2008: ‘Standard test method for determining residual stresses by the holedrilling straingage method’, ASTM International.
 Zhang Y H and Maddox, S J, 2011: ‘Development of fatigue testing of fullscale girth welded pipes under variable amplitude loading’, OMAE 30th international conference, paper OMAE2011 49942.
 Kingston E J, Stefanescu D, Mahmoudi A H, Truman C E and Smith, D J, 2006: ’Novel Applications of the DeepHole Drilling Technique for Measuring ThroughThickness Residual Stress Distributions’, Journal of ASTM International, April 2006, 3, 4, 112.
 Rosenthal D and Norton J T, 1945: ‘A method of measuring triaxial residual stresses in plates’, Welding Journal, Research Supplement, 295s307s, May.
 He W, Wei L and Smith S D (2010): ‘Intensive validation of computer prediction of residual stresses in multipass butt weld’, OMAE 29th international conference, paper OMAE201020490.
 K. Ogawa K, Chidwick L O, Kingston E J, Dennis R, Bray D and Yanagida N, 2008 ‘The Measurement and modelling of residual stresses in a stainless steel pipe girth weld’, Proceedings of PVP 2008, Paper No. PVP 200861542, July 2731, 2008, Chicago, Illinois, USA.
 Bate S, 2007: ‘Development of new residual stress profiles for BS 7910’, Serco Report Ref. No.: SA/EIG/11508/R001, Feb. 2007.
 Glinka G and Shen G, 1991: ‘Universal features of weight functions for cracks in mode I‘, Engineering Fracture Mechanics, 40 (1991) 11351146.
 Zhang Y H and Maddox S J, 2012: ’Fatigue testing of fullscale girth welded pipes under variable amplitude loading’, OMAE 31st international conference, Paper No.OMAE201283054, 16 July, 2012, Rio de Janeiro, Brazil.
 Bowman D A (2010): ‘Residual stress measurement of a 16 inch girth welded pipe using DHD technique’, Veqter report R09016b.