K. Sotoudeh, S. E. Eren, M.F. Gittos and M. Milititsky
S. Kabra and S.Y. Zhang
STFC Rutherford Appleton Laboratory
Harwell Oxford, Didcot, UK.
Paper presented at OMAE 2013 - Proceedings of the ASME 2013 32nd International Conference on Ocean, Offshore and Arctic Engineering, Nantes, France. 9-14 June 2013.
Welds between ferritic and austenitic materials, commonly described as dissimilar metal welds (DMW’s), are widely employed in various industrial applications. Failure of such joint can occur by a variety of failure mechanisms including creep, thermal fatigue and hydrogen embrittlement/corrosion While in some cases, the presence of long range stresses, due to design and fit-up, have been linked to failures, triaxial stress distribution across these types of joints have not adequately been investigated.
In the present work, forged low alloy connectors were buttered with a nickel alloy and post-weld heat treated, before making closing welds to a pipeline steel. The DMW received no further post-weld heat treatment (PWHT). The residual strain profiles across dissimilar joints were measured in the axial, radial and hoop directions of a series of these joints containing multi-run girth welds, using the neutron diffraction technique. Residual stresses were calculated from these measurements. Three dissimilar metal interface combinations were investigated: (1) ASTM A182 F22 steel to ERNiCrMo-3 (alloy 625) weld metal (PWHT’d joint), (2) AISI 8630M steel to ERNiCrMo-3 (alloy 625) weld metal (PWHT’d joint) and (3) ASTM A964-F65 low alloy steel to ERNiCrMo-3 (alloy 625) weld metal without PWHT. The current study attempts to create a better insight into the unique stress distributions across these joints.
Dissimilar metal joints are an integral part of subsea oil and gas systems. In subsea oil and gas systems, high strength steel forgings need to be welded into production systems without PWHT. This poses an engineering challenge as the hardenable forgings need to be PWHT’d to reduce their hardness following welding. To circumvent this problem, the forging material can be “buttered” with a few layers of a suitable weld metal, which can be subjected to PWHT whilst being able to accommodate an as-welded heat-affected zone (HAZ) from the subsequent completion weld. The closure weld may need to be compatible with an internal cladding material and alloy 625 has frequently been used.
In the field, there have been many years of successful service but occasional, high profile failures have been observed at the interface with the higher strength, steel [1, 2]. The fracture mechanism has been attributed to embrittlement of the microstructure by hydrogen, stemming from the cathodic polarisation of the structure for corrosion protection.
The challenge is to determine the unique conditions which have led to failure of a small number of components in the field, as the absence of a clear criterion for failure creates an uncertainty over the integrity of multi-billion pound infrastructures installed world-wide. In particular, although it is clear that in each case a crack propagated slowly to cause failure; it is not clear what the unique conditions were for crack initiation and it has been postulated that high stresses need to be present across the dissimilar metal interface for cracking to occur. However, even though in some cases long range stresses could have been inherent in particular designs, no unique stress conditions for failure have been established. This study aims at measuring the residual stresses across the following dissimilar metal interface combinations:
- ASTM A182 F22 steel to ERNiCrMo-3 (alloy 625) weld metal, PWHT’d joint, hereafter referred to as W21;
- AISI 8630M steel to ERNiCrMo-3 (alloy 625) weld metal, PWHT’d joint, hereafter referred to as W23; and
- ASTM A964-F65 low alloy steel to ERNiCrMo-3 (alloy 625) weld metal without PWHT, hereafter referred to as W25.
Table 1 – Summary of the joints
||ASTM A 182 F22
||ASTM A 964-F65
The joints are summarised in Table 1. The overall objective of the on-going project at TWI is to use the residual stress results together with small-scale and full-scale testing of components to determine the unique stress conditions for cracking.
Materials, welding procedure and specimen preparation for residual stress measurements
Table 2 – The actual chemical compositions of the three steel pipes examined.
|ASTM A 182 F22
Table 3 – The actual room-temperature mechanical properties of the three steel pipes examined.
|ASTM A 182 F22
|ASTM A 964-F65
Table 4 – Typical chemical composition of ERNiCrMo-3 nickel alloy welding consumable.
Table 5 – Typical room-temperature mechanical properties of ERNiCrMo-3 nickel alloy welding consumable.
For test purposes, welds were made between two buttered forgings of the same alloy type. All three forged pipes used for measurements had inner and outer diameters (ID and OD) of 190 and 250mm, respectively, and were 1400mm in length, with the joints positioned in the middle of the coupons, as shown in Figure 1. The chemical compositions and room-temperature mechanical properties of the steel pipes are given in Tables 2 and 3, respectively. In all cases, the multi-run girth closure weld was made to the preheated pipes using an ERNiCrMo-3 (AWS A5.14 Alloy 625) welding consumable with a width at the external surface of 41±1mm. Typical chemical composition and room-temperature mechanical properties of ERNiCrMo-3 weld metal are presented in Table 4 and Table 5, respectively.
Figure 2 – Macros of welds examined in this study
The ERNiCrMo-3 buttering was only applied to the ASTM A182 F22 and AISI 8630M steel pipes by Gas Tungsten Arc Welding (GTAW), on both 30º-beveled faces of the joint, prior to preparation for final welding. The buttering width, on either sides of the closure weld, was visually measured to be 23±1mm. PWHT of the buttered ASTM A182 F22 and AISI 8630M steel pipes was carried out at soak temperatures of 650 and 675°C and minimum soak times of 8 and 2h, respectively. The joint in the third pipe, ASTM A964-F65 low alloy steel, was made without the application of a butter layer or PWHT. Macrographs, shown in Figure 2, illustrate the three joints examined.
Figure 3 - Neutron beam paths for the measurement lattice spacing in the axial, hoop and radial directions (for dimensions of the pipe see Figure 1).
It was necessary to cut two windows through the wall of each pipe, to shorten the length of the beam path, and hence reduce counting times. The windows were cut to a sufficient size and reasonably distant from the measurement locations to: (i) provide enough room for scanning target area with the neutron beam, (ii) allow for the production of combs (or d0 samples; representative of stress-free lattice parameters) from the extracted material and (iii) minimise relaxation of the residual stresses. Neutron beam paths for the measurement of lattice spacing in axial, hoop and radial directions are shown in Figure 3.
Due to the geometry of the pipes, cutting of the windows was completed using a combination of water jet cutting and chain drilling. Residual stress relaxation during the cutting/machining processes was monitored by obtaining direct readings of linear strains from two sets of strain gauges, attached to the outer surfaces of the joints and pipes, at the 3 o’clock position (an arbitrary position chosen for neutron diffraction measurements) of the pipe. Strain readings, during different stages of the cutting/machining processes, were all compressive and did not exceed 110μ-strain in magnitude, suggesting the release of principal tensile stresses to be less than 30MPa.
Figure 4 – d0 samples prepared for the determination of stress-free lattice parameters.
The design of the combs was essentially limited to the size and locations of the materials extracted from the pipes at the window location (in the vicinity of the weld). The combs were produced to: (i) sample the materials present in each joint, i.e. the weld, buttering, parent material and the respective interfaces and (ii) meet the collimation system requirements of the actual measurement. A precise, cold machining method, i.e. electro discharge machining (EDM), using a Ø0.25mm wire, was employed to make the combs by creating a series of symmetrical 6×6×30mm teeth, allowing the complete fit of a gauge volume of 4×4×4mm. Figure 4 illustrates the comb designs and the associated details.
Measurement of residual stresses via neutron diffraction technique
Welding induced residual stresses affect the defect tolerance of welded structures. They affect the resistance to fracture because they change the crack driving force. They also affect fatigue crack propagation behaviour because they can be either tensile or compressive and therefore they can move parts of the applied stress range into or out of regimes of cyclic damage accumulation. Hence, accurate determination of residual stresses present on real structures is a vital step while assessing the integrity of welded structures [3-5].
Various methods for the measurement of residual stresses have been developed within the last thirty years. The favourable methods are certainly those considered to be non-destructive, especially diffraction-based techniques.
Figure 5 – Residual stress measurement set-up at ISIS a) Set-up for the measurement of axial strains
Figure 5 – Residual stress measurement set-up at ISIS b) Set-up for the measurement of radial and hoop strains
In this study, neutron diffraction technique was used for the measurement of residual strains on the reference, d0, samples (Figure 4) and the three welded pipes (Figure 5). Although the method itself is non-destructive, two windows were cut as described in the previous section in order to reduce the path length of neutrons and by that to reduce the measurement time. The measurement of one hoop strain component at a given location could have taken approximately 8 hours, if the window had not been cut.
For these measurements, the ENGIN-X instrument of ISIS (Rutherford Appleton Laboratory, Harwell Oxford, UK) was used. The advantage of the ENGIN-X instrument is that two orthogonal components of residual strain can be scanned at the same time with the help of the two detectors placed at both sides of the pipe (Figure 5). After the rotation of the sample, the third component of the residual strain is measured in the second set-up (Figure 5 a). For the measurement hoop and radial strain components, a two-detector set-up was used. However, for the measurement of axial strain component, only one detector was used. The second detector had to be taken out due to space limitation, to fit the pipe in the instrument in horizontal position.
The basic principle of residual strain scanning requires the determination of lattice spacing, d hkl, hkl denoting crystallographic plane. The lattice spacing can be calculated with Bragg’s diffraction law:
where, λhkl is the wavelength, θhkl is the diffraction angle, h is Planck constant, mn is the neutron mass, L is the neutron’s travel down flight path and thkl is the time of flight (ToF).
Neutron strain scanning provides the average elastic strain within the gauge volume defined by the intersection of incident and diffracted beams. After obtaining dhkl , the elastic strain, εxx, along a direction ‘x’, is determined from the change in the lattice spacing, dxx, of the crystalline material referred to the stress free value, d0, thus;
where d0 is measured on the stress-free reference samples described in the previous section.
Stress calculations following the strain calculations are based on continuum mechanics, using Hooke's law. The stress is calculated from the elastic strains in the gauge volume measured along three mutually orthogonal directions. In the case of the measurements conducted on pipes, the hoop, radial and axial residual stress components are calculated from the strains using the following equations:
where E is the Young’s modulus and ν is the Poisson ratio of the material.
The experimental uncertainties presented in this paper were derived using an error propagation method of the form:
where z is a function of x and y; and Δx and Δy are the uncertainties in x and y. The uncertainty in a strain component εxx;(say) is therefore:
if equation (2) is substituted in equation (7), which can be rewritten as:
The uncertainty in a stress component σΜ (say) is similarly derived, i.e.:
after the substitution of equation (3) into equation (9), the following equation is obtained for the uncertainty in the stress component:
Within this study, the residual strains (in hoop, radial and axial directions) were measured along two lines, parallel to the pipe longitudinal direction, sampling the joint and parent pipe. The first line was 3mm below the outer surface of the pipe and the location of the second line was 3mm above the inner surface of the pipe (Figure 3). The reason for choosing these specific locations for the measurements was related with the gauge volume chosen. The gauge volume adopted for the measurements was 4×4×4 mm. 2.82mm is the minimum distance which can ensure that the gauge volume is occupied 100%. The weld cap was left intact for the measurements. All measurements were conducted at the 3 o'clock position of the pipe.
The joints were symmetric and it would seem fair to assume that the residual stress distribution would also be more or less symmetric. Hence, the measurements were conducted starting from the weld centre and ending approximately 90mm away from the weld centre in the direction highlighted with arrows.
It was planned to measure at least 20 points on each line –biased towards the weld centre. Effort was made to measure the same or closest locations on the d0 sample.
The following measurement sequence was followed:
- measurement of stress-free lattice parameters, d0, in radial, axial and hoop directions for each weld
- measurement of axial strains on three pipes (Figure 5a)
- measurement of radial and hoop strains on three pipes (Figure 5b)
All lattice parameters were calculated using an in-house software, OpenGenie developed and maintained by ISIS, where diffracted peaks were fitted over a user-defined range.
Factors affecting accuracy of measurements
As can be seen in Figure 2, a Ni-based alloy dominates in the weld zone and the parent material is ferritic steel. When two dissimilar materials are welded, it can be expected that regions containing both iron and nickel phases will evolve. The measurement of strains in the vicinity of this interface is not a straightforward exercise. It is common practice to measure the strains in regions demonstrating a single phase and avoid focusing the beam on the regions exhibiting dual or multiple phase . However, the region of interest, which will be the determining factor in the performance of the welded structure, is in the vicinity of this interface.
Figure 6 – Peak intensity recorded in the measurement of a location exhibiting dual-phase
It is known that significant research is required for developing accurate methods to be used in the precise determination of phase fraction in a gauge volume and the contribution of these phases to the inherent strain at the location of interest. In this study, the phase fractions were estimated by examining the peak intensities, see Figure 6. The estimated values may vary ± 15 %, when such a coarse estimate is made.
Unfortunately, there is no rigorous method which allows accurate calculation of phase fractions. There were also other factors related with the quality of data, e.g. the structure of engineering materials, and the limitations of the instrument used, made the measurements more difficult or in some cases impossible. It would be useful to highlight some of these difficulties.
1 – Some data points exhibited non-optimal statistics due to the thickness of the pipe, although reasonable measurement time was allocated for each location. The lattice spacing calculated from the raw measurement data could be used for the calculation of strains, provided that their quality was within certain limits. However, a very low background signal is required for accurate determination of phase fractions, which is not easy to assure.
2 – In real engineering materials and weldments, there is usually significant texture which varies from location to location. In the case of these dissimilar joints, texture was very significant. Since the detector coverage in the sample orientation space was very limited, it was seen that some peaks dominated depending on the direction measured, whereas some were not present at all (for the same phase fraction).
3 – The beam divergence should also be noted. The beam may be more divergent in the vertical direction than the horizontal direction. This means that it may be possible to sample slightly different locations during the three measurements.
4 – Due to time constraints, a very small number of locations were measured using a 6×4×4 mm3 gauge volume. The gauge volume adopted for most of the measurements was 4×4×4 mm3. Unfortunately, the statistics of some points were still poor and the strains could not be measured even after increasing the gauge volume and allocating 3-6 hours to the measurement of each point.
In the 'Results' section, where the residual stress measurements results are presented, the measurements made at location exhibiting dual-phase in the gauge volume will be represented by three different points. One point has been calculated assuming that the dominating phase in the gauge volume is Ni, the second point has been calculated assuming that the dominating phase in the gauge volume is Fe, and the third point is the approximate, average weighted value calculated taking into account the phase fractions estimated visually from the peak intensities (see Figure 6 as an example).
It is hoped that the simplistic, pragmatic approach adopted in this study stimulates discussions in the residual stress community and a more accurate way for determining residual strains in the presence of dual or multiple phases is developed in the future. It is worth noting again that the identification of the stress state in the vicinity of the dissimilar interface is of significant importance from a structural integrity point of view. Certainly, the gauge volume could have been reduced to 1×1×1 mm3. However, the measurement of each point would have taken more than 24 hours in this case without guaranteeing small enough gauge volume. /p>
Results and observations
First of all, the distribution of residual strain -measured on three pipes and along two lines on each pipe will be discussed. This will be followed by the presentation of the residual stress distributions.
Figure 7 – Hoop, axial and radial residual strain distributions measured on the pipe designated as W25, 3mm below the outer surface.
The first pipe examined was the one designated as W25 without the application of a butter layer or PWHT. In Figure 7, the μ-strain distribution in three directions (hoop, axial and radial) versus the distance across the weld and measured along the line 3mm below the outer surface are presented. The hoop and axial strains are positive, whereas the radial strains are negative. At three locations across the weld, the presence of dual-phase was evident. These were the points 3, 15 and 18mm away from the weld centreline.
Figure 8 – Hoop, axial and radial residual strain distributions measured on the pipe designated as W25, 3mm above the inner surface
When the measurements conducted along the line 3mm above the inner surface are examined, see Figure 8, it is seen that the axial and hoop strains are negative and the radial strains are positive. The presence of a dual-phase crystallographic structure is evident at the weld centreline and 3mm away from the weld centreline.
Figure 9 – Hoop, axial and radial residual strain distributions measured on the pipe designated as W21, 3mm below the outer surface
The next set of measurements was conducted on the welded pipe designated as W21. A butter layer was applied before the closure weld and was subjected to post weld heat treatment. In Figure 9, the graphs of μ-strain versus distance across the weld (obtained from the measurements along the line 3mm below outer surface) are illustrated, where the hoop and axial strains are in general positive and the radial strains are negative. In the vicinity of the interface between the buttering layer and ferritic steel, dual-phase was observed at locations 36-42mm away from the weld centreline.
Figure 10 – Hoop, axial and radial residual strain distributions measured on the pipe designated as W21, 3mm above the inner surface
These measurements were followed by the measurements conducted on the same pipe but along another line, which was 3mm above the inner surface. In these measurements, it was seen that the hoop and axial strains were in general negative and the radial strains were positive at some locations and negative at some locations, Figure 10. Around 24-27mm away from the weld centreline, Ni and Fe-phases were present in the same gauge volume.
Figure 11 – Hoop, axial and radial residual strain distributions measured on the pipe designated as W23, 3mm below the outer surface
The pipe designated as W23 containing a butter layer and post weld heat treated was the next pipe on which the measurements were conducted. The first measurement line across this weld was again 3mm below the outer surface. As can be seen in Figure 11, the hoop strains could not be measured between 3mm and 36mm from the weld centreline. The height of the weld cap increasing the path length can be partially responsible for the elimination of the points with poor statistics. Although various attempts were made by varying the gauge volume and increasing the measurement time allocated for each point, the hoop strain distribution across this line could not be successfully obtained. The axial strains on this line were positive and the radial strains were negative. Dual-phase is present in the vicinity of 36-39mm away from the weld centreline.
Figure 12 – Hoop, axial and radial residual strain distributions measured on the pipe designated as W23, 3mm above the inner surface
Figure 12 gives the last series of measurements made on the pipe spool W23. The axial strains are negative. The hoop and radial components of strain are partially positive partially negative. Dual-phase was observed nearly 24-27mm away from the weld centreline.
Using equations (3) to (5), the three stress components acting in hoop, axial and radial directions were calculated from the obtainable, associated strain components.
Figure 13 – Hoop, axial and radial residual stress distributions measured on the pipe designated as W25, 3mm below the outer surface
In Figure 13, the residual stress distributions on W25 along the line closer to the outer surface of the pipe can be seen. The stresses in three directions show the same trend by falling to significantly low values compared to those at the weld centreline and increasing again towards the interface. The measurement points in the ferritic steel follow a steady trend. The radial stresses are in the neighbourhood of 0MPa in the parent material. Both hoop and axial stresses decreased but they did not vanish completely.
Figure 14 – Hoop, axial and radial residual stress distributions measured on the pipe designated as W25, 3mm above the inner surface
In Figure 14, the stress distributions along the line closer to the inner surface of the pipe spool W25 can be seen. The residual stresses in the root region were all of compressive nature up to 50mm away from the weld centreline. After 50mm, the axial stresses were still negative, whereas positive values of hoop and radial components were observed at some locations far away from the weld centreline.
Figure 15 – Hoop, axial and radial residual stress distributions measured on the pipe designated as W21, 3mm below the outer surface
In Figure 15, the stresses measured along the line closer to the outer surface of the pipe spool W21 can be seen. Unfortunately, the error reported for the points measured between the weld centreline and 12mm away from the weld centreline were too high and meaningful strain or stress values could not be calculated at the corresponding locations. However, it is seen clearly in this figure that the residual stresses increased towards the first interface between the closure weld and the buttering layer. In the buttering layer the hoop and axial stresses decrease but all stress components increase again towards the second interface between the buttering layer and the ferritic steel.
Figure 16 – Hoop, axial and radial residual stress distributions measured on the pipe designated as W21, 3mm above the inner surface
In Figure 16, the stress distributions along the line closer to the inner pipe surface are illustrated. These are the stress distributions closer to the root of the weld and again highly compressive stresses were observed in the root region. The stresses between 9mm and 18mm away from the weld centreline could not be measured. However, the increase in residual stresses towards the interfaces is again evident in these graphs.
Figure 17 – Hoop, axial and radial residual stress distributions measured on the pipe designated as W23, 3mm below the outer surface
Figure 18 – Hoop, axial and radial residual stress distributions measured on the pipe designated as W23, 3mm above the inner surface
In Figure 17 and Figure 18, the residual stress distributions along the lines closer to the outer and inner surfaces of the pipe designated as W23 can be seen. Along the outer line, hoop strain data could not be collected between 3mm and 36mm away from the weld centreline. Hence, the residual stresses could not be calculated. However, when Figure 17 is examined, it can be seen that the stresses increase towards the second interface between the buttering layer and the ferritic steel. Due to the lack of data points, it is not possible to draw a conclusion on the behaviour of residual stresses in the vicinity of the interface between the closure weld and the buttering layer. In Figure 18, where the stress distribution closer to the inner surface is presented, the residual stresses continuously decrease until the second interface is reached. All stress components are compressive until 75mm.
In this study, the residual stress distributions at nickel-to-steel dissimilar joints were investigated. The measurements were conducted, using the neutron diffraction technique, on three different dissimilar joints made in steel pipes. The overall objective of the on-going project at TWI was to understand the unique stress distributions across these joints, in particular at the nickel-to-steel dissimilar interface, better.
Various important observations were made on the tri-axial stress state of the as-welded pipe spools during the measurement of residual stresses. Before summarising the results, the main difficulty experienced during the residual stress measurements should be emphasised once again. As expected, at and in the vicinity of the buttering – ferritic steel interface, regions containing both iron and nickel phases were observed. However, the quantification of the residual strains at these locations or the identification of the contribution of each phase to the residual strain at those locations is not a straightforward exercise. Considerable research effort is required for the exact determination of phase fractions and the development of methods to be used in the quantification of strain at the region of interest. In this study, a practical engineering approach was adopted for this purpose, but it is certainly open to both discussion and improvement.
At the locations where the existence of dual-phase was evident, three points were plotted on the respective strain and stress diagrams. One point was calculated assuming that the material in the gauge volume is 100% nickel. For the other point, it was assumed that the material in the gauge volume consists of 100% α-phase iron. With this approach, the possible lower and upper bounds of residual stresses were estimated. The third point was plotted taking into account the phase fractions obtained from the intensities.
The most important findings noted during these measurements were that:
(i) the residual stresses in the root region of the welds were significantly lower than the stresses in the cap region and they were of compressive nature. This is definitely an advantage from the viewpoint of structural integrity. The presence of compressive residual stresses decreases the crack driving force. However, it should be borne in mind that the effect of residual stresses is very much dependent on the location of the flaw and also on the redistribution of residual stresses after the introduction or formation of the flaw. The current profiles indicate beneficial mechanical characteristics for internal surface flaws oriented circumferentially, which are one of the most frequently occurring flaw types.
(ii) the residual stresses exhibited a consistent, increasing trend towards to the interfaces in the welds with or without buttering welds. Beyond the interface, he residual stresses started to drop.
(iii) in Joints W21 and W23, the residual stresses leveled out to values below 200MPa at and around a distance of 80mm from the weld centreline. Joint W25 however, reached that value at closer distance of approximately 60mm; due to the absence of a butter layer.
The following observations were made during the residual stress measurements:
1) the residual stresses in the root region of the welds were significantly lower than the stresses in the cap region and they were of compressive nature.
2) the residual stresses exhibited a consistent, increasing trend towards to the interfaces in the welds with or without buttering welds.
3) the peak residual stress values observed in the nickel welds were in the neighbourhood of the yield strength of the welding consumable.
4) the residual stress distributions of the post weld heat treated pipe spools W21 and W23 containing buttering layer were not significantly different.
5) the tensile residual stresses measured closer to the outer surface of the pipe spool W25 (without the application of a butter layer and PWHT) exhibited slightly lower residual stresses than those in W21 and W23.
6) since there was only one interface which could locally increase residual stresses on W25, the hoop and axial components of residual stresses in the parent material started to dropp to a value in the vicinity of 200 MPa, closer to the weld centerline (approximately 60mm).
7) however, if the residual stress profiles at the root region are compared, it is seen that the residual stresses on pipe spools W21 and W23 are more compressive than the residual stresses on pipe spool W25.
Future work and outlook
Small-scale material characterisation tests, large-scale component tests, residual stress measurements using another technique (namely deep hole drilling technique), finite element modelling of the welding procedure, metallurgical examinations and structural integrity assessments are planned to be completed soon. It is believed that more conclusive statements can be made with all these data in hand.
This work was carried out under a Group Sponsored Project (GSP) of TWI funded by ExxonMobil, Statoil, Petrobras, BHP, FMC, Cameron and BP, and also under ISIS’s Industrial Collaboration R&D Scheme, funded by the Science and Technology Facilities Council (STFC).
 Burk J D and Ribardo C L, Offshore Technology Conference 2010, Paper OTC 20401.
 Kvaale P and Rorvik G, 23rd International Conference on Offshore Mechanics & Arctic Engineering 2004, Paper OMAE 2004-5136.
 BS 7910:2005, Guide on methods for assessing the susceptibility of flaws in structures, Incorporating Amendment No. 1, BSI, London, ISBN 978-0-580-60108-8, 2007.
 R6: Assessment of the Integrity of Structures Containing Defects, British Energy Generation Report R/H/R6, Revision 4, 2001.
 Glinka G., Residual Stresses in Fatigue and Fracture: Theoretical Analysis and Experiments, Advances in Surface Treatment and Residual Stresses, Ed. A. Niku-Lari, pp. 413-454, Pergamon Press, 1987.
 Skouras A., Paradowska A., Peel M.J. Flewitt P.E.J. and Pavier M.J., Residual stress measurements in ferritic steel / In625 superalloy dissimilar metal weldment using neutron diffraction and deep-hole drilling, International Journal of Pressure Vessel and Piping 101, pp. 143-153, 2013.