Quantification of residual stress effects in the SINTAP defect assessment procedure for welded components

A. Stacey, Offshore Safety Division, Health & Safety Executive, London, United Kingdom

J.-Y. Barthelemy, Institut de Soudure, Ennery, France

R. A. Ainsworth, Nuclear Electric Ltd, Barnwood, United Kingdom

R. H. Leggatt, The Welding Institute, Cambridge, United Kingdom

S. K. Bate, AEA Technology, Risley, United Kingdom  

Proceedings of OMAE99, 18 ^th International Conference on Offshore Mechanics and Arctic Engineering
July 11-16, 1999, St. Johns, Newfoundland, Canada

Abstract

The EC funded project Structural Integrity Assessment Procedures for European industry, SINTAP, has provided the opportunity to perform an extensive investigation on residual stresses and develop further the BS 7910 and Nuclear Electric R6 procedures. It has entailed an extensive literature review of residual stresses in the principal weld geometries (including plate butt, pipe butt, pipe to plate, T-butt and tubular welded joints), experimental and numerical investigations and the development and validation of procedures.

Introduction

Experimental studies have shown that secondary stresses introduced by welding or by temperature gradients can have a significant effect on the load carrying capacity of a component containing a flaw. For this reason, the SINTAP project ^[1] has included a specific task on residual stresses with the overall aim of determining and validating the most appropriate methods of accounting for residual stresses in as-welded, weld repaired and post-weld heat treated welded joints for use in structural integrity assessment. Background information and an interim report on the task on residual stresses was presented in ^[2] . In summary, the SINTAP task on residual stresses comprised a number of sub-tasks:

• Status review: Review of existing information on the treatment of residual stresses in fracture prediction, including code-defined secondary stress profiles.
  
  
• Collation of Residual Stress Profiles: Collation of existing experimental and numerically predicted residual stress profiles; comparison with distributions recommended in codes, namely the draft BS 7910 document ^[3] and R6 ^[4] .
  
  
• Experimental and Numerical Studies: Experimental and numerical investigations of residual stress distributions in welded joints. The numerical analysis work addressed the need for post-weld heat treatment (PWHT), thedetermination of its effectiveness and the derivation of criteria for PWHT of as-welded and repair-welded structures. Studies of through-wall defects were performed with reference to available experimental data and included furtherexperimental work on thick, welded A533B steel plates. In addition, a study was performed of the estimation of J-integral when dominant residual stresses are present using centre-cracked panels, thus providing an insight into theinfluence of residual stresses on the fracture behaviour.
  
  
• Standardised Residual Stress Profiles: Derivation of standardised residual stress profiles for transverse and longitudinal through thickness residual stress distributions in a range of geometries of welded jointsmanufactured from ferritic and austenitic steels; definition of a framework for incorporation of residual stresses into fracture assessment procedures based on the Failure Assessment Diagram (FAD).
  
  
• Residual Stress Intensity Factors: Derivation of residual stress intensity factors for surface cracks in a range of geometries of welded joints, i.e.
  • plate butt welds
  • T-butt and fillet welded joints
  • pipe butt welds
  • pipe seam welds
  • pipe to plate joints
  • tubular joints
  • repair welds
  
  
• Procedure Development & Validation: Development and validation of guidelines for the assessment of residual stress effects; incorporation into the SINTAP procedure.

An overview of the results from the residual stresses task are presented in this paper.

Residual stress distributions in welded joints

• plate butt welds
• pipe seam welds
• circumferential butt joints
• T-butt joints
• pipe-to-plate and tubular joints
• repair welds

The residual stress distributions in R6 ^[4] were used as the basis of the review and were supplemented by more recent data from the literature. The data are valid for the parameter ranges shown in Table 1. Further details on the through-thickness distributions are given in ^[5] . Residual stress distributions were also generated by numerical analysis for plate and pipe butt welded joints, PWHT ferritic steel joints and dissimilar metal welds ^[6] .

Table 1 Overview of residual stress data

The review was subsequently extended to include additional geometries, i.e. set in and set on nozzles, information generated in SINTAP and surface residual stresses. Generic upper bound distributions were then derived for the through-thickness and surface distributions in both the longitudinal and transverse directions in ferritic and austenitic steel.

The through-thickness distributions are presented below. They are normalised with respect to the wall thickness and the 0.2% proof stress or yield strength, σ _Y . For transverse residual stresses σ _Y should be taken to be the lesser of the yield stresses of the parent and weld material. The greater value of the two yield stresses should be used for longitudinal residual stresses and for repair welds.

The distributions are expressed either as a function of the welding conditions and of the mechanical properties of the materials or as a polynomial function. Two sets of distributions are available:

(a) for joints where the welding conditions are known
(b) for joints where the welding conditions are not known.
The distributions for (b) are more conservative than those for (a).

The surface distributions are reported in ^[7] and are not included in this paper.

Plate butt joints & pipe seam welds

The same distributions apply to both plate butt joints and pipe seam welds.

Longitudinal stresses

(a) For ferritic steels, √( πa), are plotted for infinite surface defects subject to the SINTAP polynomial profile and for residual stresses equal to the yield strength in Figure 2. The solutions are identical as crack depth tends to zero (where the residual stress equals yield in both cases), but the solution for uniform stress becomes increasingly conservative with crack depth. This demonstrates the benefits to be obtained using the SlNTAP profile, especially for deeper cracks. The solution for uniform stresses was also calculated using a closed form solution from Tada, Paris and Irwin ^[10] , which gave virtually identical results.

Fig.2. Normalised residual stress intensity factors for infinite surface defect in plate butt joint

The stress intensity factor solutions for infinite surface defects subject to the SINTAP polynomial residual stress profile and for the deepest point and surface intersection point of semi-elliptical cracks of aspect ratios, 2c/a, of 10 and 3.33 are shown in Figure 3. The stress intensity rises for a/T > 0.2 for the infinite surface crack, but falls continuously with crack depth for the deepest point of the semi-elliptical cracks. The stress intensity is always positive, despite the fact that the residual stress is negative for crack depths between 0.5 and 0.8 of the plate thickness. The stress intensity at the surface intersection point is greater than that at the deepest point for a/T > 0.5 for a crack aspect ratio of 10, and for a/T > 0.1 for aspect ratio 3.33. Hence, fracture may initiate at or near the surface intersection, particularly for the crack of aspect ratio 3.33, for which the stress intensity at the surface point is greater than that for a crack of the same depth with an aspect ratio of 10.

Fig.3. Normalised residual stress intensity factors for deepest and surface points of surface defects in plate butt joint

Treatment of residual stresses in SINTAP defect assessment procedure

where ρ is a plasticity correction factor.

An alternative approach for the assessment of plasticity effects on the residual stress distribution has been proposed in SINTAP: a factor, V, is applied to K _s to account for plasticity effects. K _r is then defined thus:

Equations relating V and ρ have been developed using previously calculated numerical values of ρ to generate values of V/V _o as a function of increasing load for a number of values of secondary stress, where

and corresponds to the value of V at zero primary stress, i.e. due to secondary stresses alone. It should be noted that the value of V becomes less than one for large L _r values corresponding to mechanical stress relief of the secondary stresses.

In ^[4] , ρ is defined as

ψ and Φ are functions of L _r and the ratio [ K _p ^s / (K _l ^p / L _r )]. Important points to note are as follows:

• for zero primary loads, ψ = 0 and Φ = 1, so that failure occurs when K _p ^s = K _mat and V = V _o .
• w is negative for large L _r values, i.e. L _r > 1.05, and this can lead to negative values of ρ when K _l ^s/K _p ^s is not significantly less than 1. Also, V < V _o. This corresponds to mechanical stress relief of the secondary stresses.
• negative values of ρ can occur even at low mechanical loads if K _p ^s < K _l ^s . This generally corresponds to plastic relaxation of secondary stresses which are elastically greater than the yield stress.
• K _p ^s can be calculated by various methods, listed below in decreasing order of complexity:
  • from equation (24) with J ^s evaluated from an elastic-plastic FE analysis;
  • from elastic-plastic analysis of the uncracked body;
  • from a plastic zone size correction to K _l ^s
  • K _p ^s can be replaced by K _l ^s when secondary stresses are low compared to the yield stress and elastic follow-up is not considered to be significant.

     (29c)

Fig.4. Numerical predictions of V/V o v. L r

Equations (28) bound the solutions for [ K _p ^s / (K _l ^p / L _r)] ≤ 2 and use a constant value for low L _r, similar to the approximate method in the current version of R6 which uses a constant value of ρ in this region. The plateau value could be made a function of the magnitude of the secondary stress. Equations (29) use a linear fit at low L _r and bounds the solutions for [ K _p ^s / (K _l ^p / L _r)] ≤ 5. At values of L _r > 0.9, equations (28) and (29) are identical and correspond to the use of ρ = 0 for L _r = 1.05 in R6. At higher values of L _r, equations (27) (28) and (29) correspond to mechanical stress relief of the secondary stresses. V/V _o has been set to 0.4 for L _r > 1.4 as a bounding value to the numerical data but, in practice, full mechanical stress relief, i.e. V = 0, has been observed experimentally ^[14] .

Validation of SINTAP residual stress assessment procedure

Numerical and experimental studies were carried out to validate the assessment procedures presented above.

Numerical studies

Additional FE analyses were performed on plates with under-, even- and over-matched welds ^[15] . The results for V are shown in Figure 5 which is similar to Figure 4, although V/V _o assumes a value of less than 0.1 (compared with 0.4 in Figure 5) for L _r > 1.05. The results confirm the suitability of equations (28) and (29). However, it should be noted that the results suggest that the proposed formulations are over-conservative for the overmatched case.

Fig.5. Comparison of equations (28) and (29) with numerical results from [15] for under-, even- and over-matched plates

Experimental investigation

In the first series of tests, five pairs of centre-cracked plate specimens, where each pair consisted of one specimen containing residual stresses and one without residual stress, were fracture tested under axial loading. The residual stress distribution was in-plane self balancing, with a uniform tensile residual stress in the central region of the plate and compression in the outer regions. The maximum residual stress was 57% of the yield strength.

The second series of tests consisted of two pairs of welded specimens, each pair consisting of an as-welded and a post weld heat treated specimen, with a semi-elliptical surface crack in the weld. PWHT entailed heat treatment of 6-10 hours at 590±10°C with a heating rate of up to 200°C/hour and furnace cooling to 200°C. The welding residual stresses were self-balancing with tension at the surfaces and compression in the central region. The as-welded and post-weld heat treated transverse residual stress distributions in the low L _r specimens are presented in Figure 6 which shows that, while the as-welded residual stresses are substantial, the residual stress levels in the post-weld heat treated specimens are relatively insignificant. The specimens were tested under four-point bending loading. One pair was tested in the low L _r regime and the other in the high L _r regime.

Fig.6. Transverse residual stress distributions in A533B-1 low L r plate butt joint specimens

The effects of residual stress on load carrying capacity with varying L _r are presented in Figure 7. The aluminium alloy specimens demonstrated the large influence of residual stresses on the load carring capacity in the low L _r region. The effect of residual stresses was found to decrease as L _r increased and had no effect at L _r and beyond, i.e. the elastic residual stress is dissipated at nett section yield. These results were confirmed by the second series of tests. The tests at low L _r showed that the load carrying capacity of the as-welded specimens was approximately 60% of that of the post weld heat treated specimens. The high L _r tests showed that the load carrying capacity was reduced in the as-welded specimen by approximately 5%.

Fig.7. Load carrying capacity ratio v. L r for aluminium test specimens

Both sets of test results are plotted against the R6 Option 1 failure assessment diagram (FAD) in Figure 8. The figure shows that, in general, conservative predictions were obtained for the aluminium alloy specimens. A small number of points fall below the curve and this was attributed to uncertainties in the experimental determination of the initiation event.

Fig.8. Comparison of aluminium residual stress test results with R6 Option 1 FAD

The predictions for the A533B-1 specimens were found to be very conservative, particularly for the as-welded specimens, even when the residual stresses were not included in the analysis. The differences were attributed to crack constraint effects, which are not considered in the R6 calculation, and to the limited number of K _mat tests. The A533B-1 steel specimens were also compared with the R6 Option 2 FAD and it was found that, at low L _r the Option 1 and Option 2 FAD predictions were very similar. However, the failure loads differed by a factor of 2 at high L _r .

Two methods were used to evaluate V:

ρ was evaluated from equation (25) with K _p ^s determined by the simplified linear elastic calculation method defined in Appendix 4 of R6 ^[4] . For the validation exercise, it was conisdered appropriate to set V _o to a value of 1, resulting in ρ being equal to ψ (see equation (26)). Values of ψ were calculated using Table A4.1 in Appendix 4 of R6. These theoretical values of ρ were compared with the corresponding experimental ρ values calculated from equation (21).

K _r was evaluated from equation (21) with L _r defined by the R6 Option 1 FAD. K _mat was evaluated from specimens which did not contain residual stresses, effectively using the R6 Option 1 FAD as a J-estimation scheme. The loads corresponding to initiation of tearing and subsequent amounts of crack growth for each specimen were converted to an L _r value from which K _r could be calculated, using the R6 Option 1 FAD. K _mat was then calculated from the relationship

Values of V from equation (30) are compared with equations (28) in Figure 9. The majority of results lie below the proposed curves and suggest that the proposed formulations for V are acceptable. The results are consistent with the discrepancies between theoretical and experimental values of ρ which are shown in Figure 10.

Fig.9. Experimentally determined V v. L r

Fig.10. Experimental ρ v. theoretical ρ

Values of V from equation (31) are compared with equations (28) in Figure 11. Except for one point, all the points lie just below the proposed formulations and the results indicate that the proposed formulations for V are appropriate.

Fig.11. V (from equation (31)) v. L r

It was concluded that the proposed formulations for V are generally acceptable as an upper bound which is independent of the magnitude of K _l ^s . Figure 11 confirms that, since ρ and V are determined from the same basis, expressing them as a multiple of rather than an additional term to ( K _l ^s/K _mat ) makes them only weakly dependent on ( K _l ^s/K _mat ).

The recommendations in BS 7910 ^[3] on (a) residual stress relaxation due to applied loading, and (b) residual stress levels in post weld heat treated specimens were also evaluated and it was concluded that the recommendations are valid.

Conclusions

• surface and through-thickness distributions have been derived for a wide range of welded plate and tubular joints
• numerical analysis has been used to predict welding residual stresses, the effects of post-weld heat treatment and the interaction of primary and secondary stresses.
• residual stress intensity factors have been evaluated for welded joints, thus providing direct quantification of residual stress effects to enable validation of assessment procedures.
• the influence of residual stresses on the fracture behaviour of welded joints has been validated experimentally.
• revised procedures for the analysis of residual stress effects in defect assessment have been developed.

Overall, SINTAP has resulted in significant advances in the characterisation and prediction of residual stress effects on the fracture behaviour of welded components. However, it has also highlighted the complex nature of residual stresses and has pointed to areas requiring further investigation. The following recommendations are made based on a review of the results generated by SINTAP:

• residual stress distributions are highly variable and there is a continuing need to obtain further information on distributions in welded joints.
• the V approach is included in SINTAP as an alternative developing method and enables V and V _o to be readily evaluated from K _p ^s , ψ and Φ which are defined in R6.
• there is a need to repeat the calculations of V/V _o v. L _r for different FADS and perform a comparison with the FE results generated in SINTAP.
• consideration should be given to reviewing the empirical relationship between VIV _o and L _r using the additional results referred to above, with a view to recommending a simple bounding curve or a table of data.
• there is a need for more extensive validation of the recommended procedure.

References

1. Webster, S. E. W., Bannister, A., SINTAP - A Structural Integrity Procedure for Europe, Proceedings of the l7th International Conference on Offshore Mechanics and Arctic Engineering, American Society of Mechanical Engineers, Lisbon, July 1998.
   
   
2. Stacey, A., & Barthelemy, J.-Y., Incorporation of Residual Stresses into the SINTAP Defect Assessment Procedure, Proceedings of the 17th International Conference on Offshore Mechanics and Arctic Engineering, American Society of Mechanical Engineers, Lisbon,July 1998.
   
   
3. BS 7910:1997, Guide on Methods for Assessing the Acceptability of Flaws in Fusion Welded Structures, Draft for Public Comment, 97/714934DC, British Standards Institution, London, 1997.
   
   
4. R/H/R6 - Revision 3, Assessment of the Integrity of Structures Containing Defects, Nuclear Electric Procedure, Nuclear Electric Ltd., Berkeley, Gloucestershire, United Kingdom, 1997.
   
   
5. Barthelemy, J. Y., Compendium of Residual Stress Profiles, Insitut de Soudure, Ennery, France, 1998.
   
   
6. Barthelemy, J. Y., Post Weld Heat Treatment of a Pipeline Butt Weld, lnsitut de Soudure, Ennery, France, 1998.
   
   
7. Leggatt, R. H, Recommendations for Revised Surface Residual Stress Profiles, Report SINTAP/TWI/4-3, The Welding Institute, Cambridge, UK, 1998.
   
   
8. SYSWELD, Instruction Manual, Version 231 , Framasoft & CSI, 1991.
   
   
9. Al Laham, S., Stress Intensity Factor and Limit Load Handbook, Report EPD/GEN/REP/O316/98, Issue 2, Nuclear Electric Ltd., Berkeley, Gloucestershire, United Kingdom, 1998.
   
   
10. Tada, H., Paris, P. C., &Irwin, G. R., The Stress Analysis of Cracks Handbook, Del Research Corporation, Hellertown, Pennsylvania, USA, 1973.
    
    
11. Newman, J. C., and Raju, I. S., Analysis of Surface Cracks in Finite Plates Under Tension or Bending Loads, NASA Technical Paper 1578, December 1979.
    
    
12. RG-CODE , Nuclear Electric plc, Berkeley, Gloucestershire, United Kingdom.
    
    
13. Ainsworth, R. A., The Treatment of Thermal and Residual Stresses in Fracture Assessments, Engineering Fracture Mechanics, Vol. 24, pp. 65-76, 1986.
    
    
14. France, C. C., Sharples, J. K., and Wignall, C., Experimental Programme to Assess the Influence of Residual Stresses on Fracture Behaviour - Summary Report, Report AEAT-4236, AEA Technology, Risley, 1998.
    
    
15. Smith, S. D., Comparison of the PD6493:1991 Rho ( ρ) Factor with FEA Results, Report SINTAP/TWI/1-2, TWI, Abington, Cambridge, 1997.