Residual stress characterization of a fabrication weld from the VICTORIA-Class submarine pressure hull : Revealing the unseen

Publisher’s version / Version de l'éditeur:

Canadian Journal of Physics, 88, October 10, pp. 759-770, 2010-10-01

READ THESE TERMS AND CONDITIONS CAREFULLY BEFORE USING THIS WEBSITE. https://nrc-publications.canada.ca/eng/copyright

Vous avez des questions? Nous pouvons vous aider. Pour communiquer directement avec un auteur, consultez la première page de la revue dans laquelle son article a été publié afin de trouver ses coordonnées. Si vous n’arrivez pas à les repérer, communiquez avec nous à PublicationsArchive-ArchivesPublications@nrc-cnrc.gc.ca.

Questions? Contact the NRC Publications Archive team at

PublicationsArchive-ArchivesPublications@nrc-cnrc.gc.ca. If you wish to email the authors directly, please see the first page of the publication for their contact information.

NRC Publications Archive

Archives des publications du CNRC

This publication could be one of several versions: author’s original, accepted manuscript or the publisher’s version. / La version de cette publication peut être l’une des suivantes : la version prépublication de l’auteur, la version acceptée du manuscrit ou la version de l’éditeur.

For the publisher’s version, please access the DOI link below./ Pour consulter la version de l’éditeur, utilisez le lien DOI ci-dessous.

https://doi.org/10.1139/P10-076

Access and use of this website and the material on it are subject to the Terms and Conditions set forth at

Residual stress characterization of a fabrication weld from the

VICTORIA-Class submarine pressure hull : Revealing the unseen

McGregor, R. J.; Rogge, R. B.

https://publications-cnrc.canada.ca/fra/droits

L’accès à ce site Web et l’utilisation de son contenu sont assujettis aux conditions présentées dans le site LISEZ CES CONDITIONS ATTENTIVEMENT AVANT D’UTILISER CE SITE WEB.

NRC Publications Record / Notice d'Archives des publications de CNRC:

Residual stress characterization of a fabrication

weld from the VICTORIA-Class submarine

pressure hull: revealing the Unseen

1

Abstract: Explicit understanding of the residual-stress character of primary submarine pressure hull weldments will

im-prove the fidelity of numerical analysis and experimentation supporting operational envelope and design life. A length of circumferential-seam closure weld was contained within a section of hull plate removed from the HMCS VICTORIA dur-ing the extended dockdur-ing work period (EDWP) refit operations. This has provided a rare opportunity for detailed charac-terization of the as-received condition of this common weld-type from original vessel assembly. In collaboration with the Canadian Neutron Beam Centre of the National Research Council (NRC), a program was conducted to study this weld us-ing neutron diffraction. Neutron diffraction is able to survey nondestructively through the section thickness, providus-ing a three-dimensional characterization, while leaving the specimen intact for complementary study by other methods. Results indicate tensile stress peaks of up to 80% of the base-material yield stress. Understanding the three-dimensional behaviour of residual stress in this type of weld provides a valuable resource to the numerical modelling community. The results can also support fatigue and fracture experimental work and serve to confirm and improve the interpretation of the existing body of ‘‘surface-only’’ work conducted on similar welds.

PACS No: 61.05.F–

Re´sume´ : Une compre´hension explicite de la nature de la charge re´siduelle de la coque pressurise´e primaire d’un

sous-ma-rin assemble´e par soudure va ame´liorer la qualite´ de l’analyse nume´rique et expe´rimentale visant a` ame´liorer sa vie ope´ra-tionnelle. Une longueur de soudure d’un joint en circonfe´rence faisait partie d’une section de la coque pre´leve´e sur le HMCS Victoria pendant les ope´rations de re´paration extensives de longue dure´e (EDWP). Ceci nous a fourni une rare op-portunite´ pour caracte´riser en de´tail la condition telle que rec¸ue de ce type de soudure commune dans la construction origi-nale du vaisseau. En collaboration avec le Centre Canadien de faisceaux de neutrons du Conseil National de Recherche, nous avons e´tudie´ cette soudure par diffusion de neutrons. La diffusion de neutron a pu sonder de fac¸on non destructive toute l’e´paisseur de la section, fournissant une image tridimensionnelle, tout en laissant le spe´cimen intact pour des analy-ses comple´mentaires utilisant d’autres me´thodes. Les re´sultats indiquent des efforts sous traction jusqu’a` 80% de la limite d’e´lasticite´ du mate´riau de base. La compre´hension du comportement tridimensionnel de la charge re´siduelle dans ce type de soudure fournit de l’information importante a` ceux qui mode´lisent nume´riquement ces structures. Les re´sultats peuvent justifier de pousser des travaux expe´rimentaux sur la fatigue et la rupture et servent a` confirmer et a` ame´liorer les me´tho-des d’examen en surface seulement que l’on fait sur me´tho-des soudures du meˆme type.

[Traduit par la Re´daction] 1. Introduction

Engineering efforts to establish and improve ultimate sub-marine operational and design-life limits will require de-tailed understanding of the physical character of the pressure hull. The residual stress character in cold rolled, cold worked, and welded regions are variables required for

analysis of various degradation mechanisms associated with the aggressive conditions experienced by a military submer-sible (buckling, fatigue, stress corrosion cracking). In situ methods have been employed by DRDC and others, past and present, in an effort to measure such residual stresses [1–5]. Results have confirmed levels of tensile stress on or near surfaces, but application of these methods is generally impeded by issues of physical access, disruption to vessel manufacture and mission operations, and the inability to ob-tain through-thickness information from available techni-ques, since in-service equipment must remain intact.

Recently a section of pressure hull plate was removed from service on HMCS VICTORIA 3_{during repair and refit}

operations at Canadian Forces Base (CFB) Esquimalt (Fig. 1). This plate was provided to DRDC Atlantic, Dock-yard Laboratory Pacific (DLP) for metallurgical research. The harvested section of hull plate contained several fabrica-tion features, including a secfabrica-tion of circumferential-seam (circ-seam) weld, one of several used to join together unit sections at fabrication.

Received 15 July 2010. Accepted 3 September 2010. Published on the NRC Research Press Web site at cjp.nrc.ca on 19 October 2010.

R.J. McGregor.2_{Defence Research and Development Canada}

-Atlantic, Dockyard Laboratory Pacific, Building D199, Victoria, BC V9A 7N2, Canada.

R.B. Rogge. National Research Council, Canadian Neutron

Beam Centre, Building 459, Station 18, Chalk River, ON K0J 1J0, Canada.

1_{Special issue on Neutron Scattering in Canada.}

2_{Corresponding author (e-mail: rod.mcgregor@forces.gc.ca).} 3_{Formerly HMS Unseen of the British Royal Navy.}

In collaboration with the NRC Canadian Neutron Beam Centre, a program was conducted to determine the residual stress character within the plate and weld. The experiment was performed over a four week period with DLP personnel supporting on-site at the L3 neutron diffractometer in Chalk River, Ontario.

The extracted plate provides a very rare opportunity for full-thickness residual stress characterization of a legacy fabrication weld from an in-service submarine pressure hull. The results can be considered representative of all circ-seam legacy welds on the VICTORIA-class pressure hull and can be used in the mechanical analysis supporting operational limits for the vessels, including diving depth. Characteriza-tion data can also support design life and (or) extension analysis and experimentation.

The plate provides the first opportunity to exploit neutron diffraction (ND) for through-thickness weld characterization on a Canadian submarine hull. Since it is nondestructive, ND avoids stress relaxation correction errors inherent in the X-ray diffraction (XRD) process when layer removal is em-ployed and leaves the specimen intact for complementary evaluation by XRD. Use of complementary methods for the same characterization objective will ensure optimal exploita-tion of the plate and validate or improve DRDC’s techniques in XRD, strain gauging, experimental fracture mechanics, and numerical simulation for submarine pressure hull opera-tions support [6, 7]. Results can also complement and im-prove the interpretation of the historical database of these types of measurements made using surface-only techniques on similar weldments.

2. The neutron diffraction method

Neutrons are electrically neutral subatomic particles hav-ing wavelike properties with a characteristic wavelength that is related to the neutron energy. Neutrons diffract when length scales in the sample are comparable to the neutron de Broglie wavelength. In the case of crystalline materials, the spacing between planes of atoms, the d-spacing, meets this requirement.

When an initially stress-free crystalline material (in-cluding a polycrystalline aggregate) is subject to a force (stress), these planes of atoms either move further apart (ten-sile stress) or move closer together (compressive stress). In a diffraction experiment, the d-spacing can be determined from the diffraction angle (2q) and the wavelength of the in-cident radiation using Bragg’s law, viz.

l_{¼ 2dsinðqÞ ð1Þ}

As such, changes in the interatomic spacing, such as those caused by stress, result in a 2q shift of the diffraction peaks. The lattice strain, e, is the fractional change in the lattice spacing with reference to the stress-free lattice spacing, do,

given by the following relationship, 3_¼

d_{ do} do

ð2Þ The strain-free lattice spacing, required for this equation, was determined from small coupons that were extracted using EDM from both a relatively stress-free section of the plate and the weld centre. Gently cutting a sufficiently small coupon mechanically relieves the residual stresses.

By definition, the scattering vector is the change in the momentum vectors of the neutron before and after interac-tion with the sample. Therefore, when elastic scattering (dif-fraction) occurs, two conditions are true: (i) the scattering vector is parallel to the bisector of the incident and scattered neutrons, and (ii) the scattering vector is perpendicular to the diffracting planes of atoms. Consequently the strain, the relative change in the perpendicular spacing between the planes of atoms, is measured in the sample direction that lies parallel to the bisector of the incident and scattered neu-tron beams.

For a weld, these directions are typically transverse (T), longitudinal (L), and surface-normal (N), as further de-scribed in the next section. Stress and strain are tensors; by measuring the normal strains in three mutually orthogonal directions the stress can be determined along those same di-rections using the generalized Hooke’s law. That is, sxfor x

= T, L, or N is given by the following equation, where E is the Young modulus and n is the Poisson ratio for the mate-rial, sx¼ E 1_{þ n} 3xþ n 1_{ 2n}ð3Tþ 3Lþ 3NÞ h i ð3Þ The experimental uncertainty in stress is calculated as fol-lows: dsx¼ E 1_{þ n} ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 1_{ 2n}d3x 2 þ n 1_{ 2n}  2 ðd3T2þ d3L2þ d3N2Þ   s ð4Þ where de is the uncertainty in strain, which in turn is ulti-mately linked to the uncertainty in the determination of the peak position. The uncertainty in peak position is deter-mined from the statistical quality of fit of the diffraction peaks to a function consisting of Gaussian profile and a background term. Further descriptions of the technique are well documented in the literature [8].

Fig. 1. HMCS VICTORIA in CFB Esquimalt drydock for

sched-uled maintenance work period. (Reproduced with permission from the Department of National Defence.)

3. Experimental approach

3.1 As-formed hull plate and circ-seam weld coupons

To provide a baseline comparison for the circ-seam weld stresses, the residual stress of a hull-plate sample from a mast casting removal–replacement program was character-ized by neutron diffraction. This as-formed plate coupon is free of the contaminating influence of fabrication welding or in-service impact. It therefore represents only the cold work of mill fabrication and cold-forming to the submarine hull design radius.

The circ-seam weld exists on a dented section of plate, which was removed by flame cutting. Figure 2 is a photomi-crograph obtained after cutting, polishing, and nitrol etching of a sample of the weld cross-section. The primary objective of this study was to characterize the through-thickness resid-ual stress distribution across an area similar to that shown in this photomicrograph. Figure 3 illustrates the stress direction conventions and their relation to physical features on the coupon and pressure hull. Figure 4 shows a photo of the outer surface of the plate, where the path of the circ-seam weld has been revealed by a nitrol etch. Figures 3 and 4 il-lustrate the challenges to select measurement planes isolated from stress-polluting influences of lug weld, flame cut, and dent.

3.2 Stress direction convention

As noted in Sect. 2, the stress in three mutually orthogo-nal directions is determined from the strains measured in those same three directions. The directions were chosen to align with sample symmetry directions. For the as-formed plate (baseline) coupon, the symmetry directions are referred to as axial and circumferential (or rolling direction) to avoid confusion with weld-oriented terminology.

For the circ-seam plate coupon, the three principal strain components are referenced to the circ-seam weld direction and plate as shown in Fig. 3. Applying the typical conven-tion for butt welds, the transverse (T) direcconven-tion is perpendic-ular to the weld bead (across the weld), the longitudinal (L) is in the direction of the weld bead (along the weld), and the surface-normal (N) is into the plate thickness (also called the radial direction relative to the hull centerline). For both plates, the transverse and longitudinal directions correspond, respectively, to the axial and circumferential directions of the submarine hull.

In neither case is it assumed that these are the principal stress directions. Determining the principal stress directions requires measuring the full strain tensor at each location. However, the principal stresses are normally aligned with the symmetry directions, and in the case of welds, the prin-cipal stress directions usually align with the weld direction and with directions transverse to the weld.

3.3 Coupon measurement locations

Baseline measurements were taken on the coupon of as-formed hull plate, extracted from a separate location on the hull. The three strain directions on this sample were selected relative to the visible curvature of the plate (also the plate-rolling direction). Nine measurements were made in each of the three directions in a line through the as-formed plate thickness from the OD to the ID, starting at a level around

3.5% of design thickness. The line location was selected at the centre of the plate coupon to minimize any stress-pollut-ing influence from the flame cuttstress-pollut-ing and (or) machinstress-pollut-ing near the edges.

The circ-seam weld coupon was initially aligned in the test frame in an orientation appropriate to position the neu-tron gauge transverse to the weld centreline at Planes 1 and 2. Because the plate is not flat or uniform in thickness, a measurement grid was established based on dial gauge char-acterization of the OD and ID surfaces at the measurement planes (Fig. 5). The area of the measurement planes was surveyed at uniformly incremented depths relative to the curved OD surface. A higher grid (measurement) density was established closer to the weld centreline.

Fig. 2. Photomicrograph of the circ-seam weld showing double-vee

weld prep design, fusion line, heat affected zone, thickness deple-tion by surface grinding, and evidence of ID back gouge.

Fig. 3. Orientation and spatial conventions used for the circ-seam

weld plate surveys. The location of the neutron diffraction mea-surement planes are indicated as Plane 1 and Plane 2.

To access the full thickness and to orient the test gauge along each of the strain directions, the plate required three positionings and precise alignments. Post-processing of the alignment data confirmed accurate co-location of the neu-tron strain gauge at the same grid locations for each of the three principal stress directions.

At each orientation of the plate, the surface profile of the measurement Planes 1 and 2 was measured by translating X,

Y, and Z of the sample table over a mounted dial gauge. Subsequent ‘‘wall scans’’, made by neutron beams at fiducial points, together with dial guage readings were used to trans-form the sample coordinate system into the lab frame so that measurement locations in the sample could be programmed into the L3 control system.

3.4 NRC CNBC L3 configuration and calibrations

For this experiment, the (115) planes of a Germanium monochromator selected neutrons of wavelength 1.567716 A˚ at a take off angle 2qm = 928. The nominal

gauge volume was set at 2  2 mm in the scattering plane, as defined by the incident and scattered beams, and 20 mm perpendicular to the plane for the N and T components and 2 mm for the L component. A computer-controlled transla-tion and rotatransla-tion system positransla-tioned the sampling volume at locations of interest within the specimen. Figure 6 shows a photo of the L3 spectrometer with the circ-seam plate speci-men aligned on the sample table for measurespeci-ment of the ra-dial component of strain.

The diffraction peak is measured using a position-sensi-tive detector and recorded as neutron counts versus diffrac-tion angle 2q. The peak posidiffrac-tion 2q, the integrated intensity

I, and the line width u were obtained by fitting the data to a function consisting of a Gaussian distribution plus a sloping background.

The diffraction angle was measured on the nominally zero stress coupons. The fusion zone shows a distinctly different

d₀spacing. All calculations of strain within the fusion zone account for this change . The wavelength of the neutron beam was calibrated with a Ni powder standard.

4. Discussion of results

4.1 As-formed plate coupon

Calculated residual stresses based on ND strain measure-ments on the baseline specimen are shown in Fig. 7, with XRD results included from [6]. The rolling (circumferential) direction exhibits the highest magnitude and variation with depth, which stems from two phases of its fabrication: the mill rolling process and the cold work of the rolling–crimp-ing formation to the hulls design radius. The distribution of stresses is typical of the equilibrated pattern of a cold bent and released plate [9]. Peak tensile and compressive stress levels nearing 200 MPa are indicated for both the hull cir-cumferential (rolling) and axial directions. This is less than 36% of 0.2% offset yield stress (sy). For both stress

direc-tions, however, neither OD nor ID surfaces exhibit the unde-sirable condition of having significant tensile stresses, which may otherwise present concern for corrosion mechanisms or reduction of structural margin during operational loading.

The XRD measurements are closer to the surface than the the ND results. They are also from a different hull plate and an independent location. The ID measurements included are the deepest XRD subsurface penetration attempted in the study, at 2.5% design wall thickness penetration from the ID [6]. Extrapolation of the ND results to the surface sug-gests good agreement with the XRD data obtained at the surfaces. Near the ID, there are XRD data obtained at a depth of 2.5% of the design thickness, which is in close proximity to the ND data at 94% from the OD. At these lo-cations we observe that stresses are similar in value. The cir-cumferential stresses are the same within uncertainty, while the axial stresses are comparable but not the same within uncertainty. However, near the surfaces the stress gradients are steep, the uncertainty in positioning and variations in thickness are therefore important. The ND-determined ra-dial-direction stress is zero within uncertainty, as it would also be for the XRD data. The trends in the ND stress data, as one approaches the surfaces where XRD data are avail-able, indicate good agreement between the complementary techniques.

Figure 8 contains ‘‘baseline’’ stress data from three inde-pendent plates. The as-formed coupon from Fig. 7 (thick line labelled ‘‘Baseline Sample’’) is depicted in combination with the near-surface XRD data from elsewhere on the sub-marine hull. Data from the circ-seam dented plate (shown as thin dashed lines) are from the peripheral extent of the circ-weld measurement planes, away from influence of the circ-weld. Also shown in the figures are the measurements at the cen-treline of the circ-weld (solid lines), which will be discussed in the next section.

The stress profiles from the dented sample are taken at Plane 1 and 2 locations and have similar characters. The scatter is typical of experimental stress measurements of ma-terial in the as-fabricated condition and is a function of measurement uncertainty and subtle variation of mechanical properties of the material. Of note, however, is the dissimi-larity to the residual stress profile from the as-formed base-line sample. Confirmation of the as-formed profile is supported by the correlation between the ND and XRD data from separate locations. This suggests that the difference be-tween the as-formed baseline profile and dented plate stress

Fig. 4. Photograph of the plate OD showing the circ-weld path

(vertical) revealed by etching, and proximity of lug weld, flame cut edges, and dent (upper right).

profiles is largely attributable to cold work and residual stress caused by the dent.

An exploratory study was conducted of the OD surface magnitude of residual stress from the dent alone. Using the finite element model (FEM) developed for [10], the dent stresses at the OD surface are around 200 MPa tensile in the hull circumferential and 50 MPa tensile in the hull axial direction. This confirms the expectation of contamination from the dent stress, particularly in the axial stress direction.

4.2 Circ-seam plate coupon

The graphs in Fig. 9 compare the final residual stress dis-tributions on lines across the weld (perpendicular to the weld bead) at incremental depths from the OD at Plane 1. In all cases, the stress character and magnitudes are similar for the tangential and longitudinal directions. Stress peaks that seem to vary with the position of the weld fusion line – heat affected zone (HAZ) are also evident.

In the first graph of the Fig. 9 series, the near OD surface stresses are plotted for both principal directions. Experimen-tal uncertainties are also shown for this data. The uncertain-ties are typical of all measurements presented and reflect a high level of confidence in the precision of the measure-ments. The magnitude and character of the stresses in Plane 1 will be the focus of discussion since they were found to be very similar to that of Plane 2 (not shown). For both planes, there is a clear and significant elevation of tensile stresses in the weld region. The peak stresses measured at the depth 3.5% of design wall thickness below OD are the highest ten-sile stresses measured with transverse stresses of 450 MPa (80% sy) and longitudinal stresses of 300 MPa (55% sy) at

Plane 1. The highest stresses are in the transverse direction and in the centre of the weld metal. For both directions and Planes, shoulder peaks occur at the location of the welds fu-sion line – heat effected zone. The character of the distribu-tions, including peak location, is typical of the observations in similar studies [9, 11]. There is no significant difference in dent influence between the more proximate (Plane 1) and distant (Plane 2) stations, although a slight trend to higher tensile magnitudes in Plane 1 may be attributable to this in-fluence. Similarly, there is also no obvious effect from prox-imity to the lug weld. However, proxprox-imity to the flame cut surface may be the source of the tensile slope increase at Y = –85 mm (less than 30 mm from the edge). Overall, the stress character similarity and the symmetry at the principal area of interest suggest only minor contamination from the welding and flame cutting.

XRD data from the work described in [6] is shown for reference. Relative to the current study, the XRD measure-ments were taken on the in situ hull at a shallower depth within the steep near-surface stress gradient; clearly, close magnitude correlation to the present study was not expected. Although the difference in location of measurement is a pri-mary factor, there will be strain contributions from weld dis-tortion relaxation at extraction of the circ-seam plate sample and dent imposed contamination to the circ-seam plate.

Fig. 5. Scan grid for transverse stress direction at Plane 1, showing dial gauge surface measurements normalized to diffractometer wall

scans.

Comparisons have been made between the stress peak measurements of this study and those from the literature on similar weldments. Figure 10 shows the curve by Yurioka et al. [9], which shows the experimentally derived relationship of the maximum residual stress as a function of yield strength of the base metal for high-strength steel butt-welds. Maximum residual stresses prove to be consistently lower than elastic yield. Confidence in the results from the current study is gained from the consistency with this relationship, even considering uncertainties in ND or sy measurements

and the contaminating influence of dent stresses.

Figure 11 shows the effect of tensioning the longitudinal stresses on this type of weld, which is significant to the in-terpretation of the results of this study. Similar to the profile change between Curve 0 and Curve 1 in Fig. 11, the stress magnitudes in the central weld region (+/–20 mm) will be less affected by the dent strain energy than the outlying areas (+/–80 mm regions). This observation is supported by the unexpectedly high stress levels at the +/–80 mm extrem-ities of Fig. 9. This indicates that there may be only minor pollution to the weld stresses from the dent and its subse-quent extraction. This nonlinear superposition effect serves

Fig. 7. Through-thickness baseline data from as-formed plate coupon (non-dented, non-welded). Surface data by X-ray diffraction are

in-cluded for comparison.

to legitimize the measured longitudinal direction stresses where NLFEA performed for this study indicates nontrivial dent imposed stresses of 200 MPa at the measurement planes (50 MPa in the transverse direction). The accuracy of the longitudinal results is further confirmed by the corre-lation observed in Fig. 12, comparing the longitudinal

stresses of Fig. 9 (this study) to both literature NLFEA and X-ray diffraction experiments on near identical weldments.

An NLFEA follow-on study could more explicitly examine the various magnitudes of influence and affect necessary cor-rections to the ND measured data to provide a more accurate indication of typical in-service residential stress. However,

based on the above assessment, such a refinement would be expected to confirm the dent cold work is of marginal im-pact. The ND measured results are considered representative to conservative, relative to a non-dented in situ hull section.

Although subject to aggressive transient loading by deep diving, compressive strains are predominant on an externally loaded pressure vessel and are expected at this specific loca-tion on the hull. As evidenced in Fig. 10, the residual stresses are not provided the benefit of the permanent me-chanical stress relief illustrated in Fig. 11.

Figure 13 compares the residual stress distribution in a line through-thickness, approximately 16 mm off the weld centreline. This station corresponds to the weld fusion lines at the OD and ID, as shown in the figure. In addition to the weld centreline, significant stress peaks are associated with this location. The ND distributions are artificially completed by inclusion of the fusion line XRD data point from [6] (Farrell and Bayley), with the ID data point as-sumed from the measured OD point based on geometric symmetry and matching fabrication steps (surface weld grinding, etc.).

Figure 13 allows direct comparison of the residual stress distribution along the line beneath the ID HAZ (tensile peak) with an available numerical prediction, as studied by Kanninen et al. [12], using similar material and weld type. Kanninen identifies this ID location as a potential crack ini-tiation site. The figure shows a comparable distribution be-tween their results and those of the current experiment. These results confirm the hypothesized presence of a steep stress gradient at the ID and OD surfaces. The experimen-tally determined gradient at the surfaces is more pro-nounced, influenced by the cold work of plate rolling and weld grinding, both of which are not addressed by Kanni-nen. Although the experiments may suggest an even greater

vulnerability based on subsurface tensile magnitudes, high compressive stresses on the exposed surfaces may afford protection from crack initiation.

4.3 Residual stress maps and visualizations

Figure 14 shows the stress data presented as isobaric (‘‘contour’’) maps using the lab adjusted coordinate conven-tion established with Fig. 5. Various stress visualizaconven-tion methods are included. In the top figure of each, lines of con-stant stress are overlaid on the photomicrograph of the circ-seam weld. Through the different visualization methods de-scribed in this section, it is interesting to compare the ND measured stress results (pattern and magnitude) with this photo. An improved comprehension is provided to the linear style of data depiction in Fig. 9.

The centre and bottom image of each figure are different presentations of the same data. In the centre, the contours are shaded dark to light indicating compressive to tensile stresses. In the bottom image, a photolike image is rendered showing similarly depicted residual stress in the dark to light convention.

In all images the shape of the weld is apparent in the stress distribution. Compared with the photomicrograph of the actual weld, a number of observations are apparent. Firstly, the stress maxima at the weld centre of the OD com-prises a shallow layer at the surface (<10% of design wall

Fig. 10. Peak residual stresses on the circ-seam weld plotted with

curve (Yurioka) showing the relationship of peak residual stress to base metal yield stress for high-strength steels [9].

Fig. 11. Effects of superposed elastic stress on weld residual stress:

1) neutralizing of residual stress profile with tensile loading, 2) re-latively small influence on peak tensile stress, 3) permanent stress relief post release [9].

Fig. 12. ND determined longitudinal component of stress across circ-weld, compared with results from highly similar weldments [9].

Fig. 13. Through-thickness transverse stress measurement by both ND and XRD for comparison with Kanninen ID to OD ‘‘potential crack

line’’ FEA prediction of residual stress on similar weldment. Comparison confirms experimental distribution of stress, presence of steep gradient at surfaces [12].

thickness) and are found in both the transverse and longitu-dinal directions. Evidently, considerably higher stress peaks at this critical location would have existed in the unfinished weld crown and have been ground away, suggesting manu-facturing efforts to temporarily extend the V and crown depth of the undressed weld may be a strategy to reduce such peaks. Secondly, this surface stress peak is not evident at the ID, which is favourable for an externally loaded pres-sure vessel. Observing the photomicrograph, it is evident from the bead pattern that an ID back gouge step was taken at manufacture. This appears to have had an ameliorating ef-fect on the resultant residual stresses at the weld centre of this surface. Thirdly, comparison of images with the photo-micrograph clearly show a stress peak effect at the HAZ of the weld, which has been reported elsewhere in the litera-ture. This has not been similarly attenuated by the back gouge. Finally, the surface-tending tensile-stresses of the double-vee butt-weld are balanced by the formation of a highly compressive region in the wall mid-section, correlat-ing in depth with the extent of the back gouge penetration.

The ability to adjust perspective with these orthographic depictions has been developed by this study to provide a novel way of visualizing residual stress using the technique of stereogrammetry. Using traditional stereoscope viewer or ‘‘cross-eye’’ technique, this type of visualization provides

3D visualization of the isobaric information. Figure 15 is a visual representation of transverse stress over the circ-weld area at Plane 2. To establish the 3D effect, the viewer crosses eyes to place the + in the O, holds it in the O for several seconds. While maintaining crossed-eyes, the viewer slowly glances down at the centre image of the three visible. The brain will ‘‘lock on’’ and find focus on the 3D image. In this depiction, tensile stress varies from light to dark red with increasing magnitude, while compressive stress varies from light to dark blue with increasing magnitude.

Observing Fig. 15 in this way, most of the information gleaned from the previous depictions becomes apparent; the shallow-depth stress maxima at the OD, the stress relief from the ID backgouge, the stress peaking at the HAZ, the compressive region in the plate centre, the similarity of transverse to longitudinal distributions, and the through-thickness character of the as-formed–dented plate away from the weld.

5. Conclusions

Neutron diffraction residual stress measurements were ac-quired on extracted samples of VCS pressure hull. These measurements allowed a through-thickness residual stress characterization of two independent planes over the

section of a circ-seam double-vee butt weld from the origi-nal manufacture. A separate coupon from an as-rolled– formed hull plate was similarly studied to provide baseline comparison from a non-dented and unwelded sample. Re-sults will be used to enhance the fidelity of present and fu-ture numerical models used to study the design life and operational envelope.

The primary conclusions of this study are as follows: 1. The near surface results from the as-formed (baseline)

plate sample, correlate to earlier DRDC X-ray diffraction studies, indicating a steep stress gradient toward the ID and OD surfaces. As-formed stress maxima are internal, with no dominance to tension or compression, or domi-nance to transverse or longitudinal direction relative to the circ-seam weld. Peak magnitudes of around 200 MPa (36% sy) were measured in the tensile and

compressive directions.

2. Comparison of XRD and ND of the circ-weld region in-dicate a steep stress gradient from high surface compres-sion to subsurface tencompres-sion within 5% of the OD or ID

surface. The presence of this near-surface gradient is confirmed by the results of numerical analysis performed by others on similar weldments. The subsurface numeri-cal analysis also confirms the distribution of the stresses through the thickness at the HAZ. Higher levels of sur-face compressive stresses are indicated by the experi-ments; these are predominantly influenced by weld finishing by grinding, and are generally considered pro-tective against insipient cracking.

3. Subject to further analytical accounting for the dent and refinement for the stress influence of residential con-straint, the residual stresses measured are representative (to somewhat conservative) of all circ-seam welds on all pressure hulls within the VICTORIA-class fleet.

4. Over the circ-weld region there is a clear and significant elevation of tensile stresses in the weld region relative to the as-formed plate. The peak stresses measured at the depth of 3.5% of design wall thickness below OD are the highest tensile stresses measured with transverse stresses of 450 MPa (80% sy) and longitudinal stresses Fig. 15. 3D stereogram residual stress visualization.

of 300 MPa (55% sy) at Plane 1. This maximum is in the

transverse direction and in the centre of the weld pool. For both principal stress directions and measured planes, shoulder peaks occur at the location of the fusion lines and heat effected zones of the welds.

5. The stress maxima at the weld centreline on the OD comprises a shallow layer at the surface (<10% of design wall thickness) and are found in both the transverse and longitudinal directions. Evidently, considerably higher stress peaks at this critical location would have existed in the unfinished weld crown and have been ground away. This suggests a manufacturing strategy to reduce such peaks by biasing high residual stresses into an exag-gerated crown, which is then removed by grinding. 6. The weld centreline transverse tensile stress-peak is not

evident at the weld ID surface, which is favourable for an externally loaded pressure vessel. An ID back gouge step was taken at manufacture, which appears to have had an ameliorating effect on the resultant residual stres-ses at the weld centre of this surface. However, trans-verse stresses reach high tensile levels on the HAZ of the ID surface. These have not been similarly attenuated by the back gouge. Tensile stresses around 65% syexist

near the ID HAZ for two of measured stress directions. References

1. J.F. Porter. Residual Stress Evaluation of HMCS ONAN-DAGA Pressure Hull Weldments, Canadian Forces/CRAD Meeting, DREA – Halifax, May 1985.

2. J.F. Porter, M.F. Brauss, and J. Pineault. X-ray diffraction determination of residential strain for marine platforms, 2nd Canadian Forces/CRAD Meeting on Naval Applications of Materials Technology, DREA-Halifax, May 1985.

3. M. Brauss, H. Wong, R. Holt, and P, Fryzuk.

Non-destruc-tive evaluation of oberon submarine pressure hull residual stress. 2nd Canadian Forces/CRAD Meeting on Research In Fabrication and Inspection of Submarine Pressure Hulls, DREA-Halifax, May 1987.

4. C. Adams and D. Corrigan. Mechanical and metallurgical behaviour of restrained welds in submarine steels. MIT Welding Lab, Defence Technical Information Centre, AD0634747, May 1966.

5. S. Khan, D. Saunders, J. Baldwin, and D. Sanford. Exp. Mech. 37, 264 (1997). doi:10.1007/BF02317417.

6. S. Farrell and C. Bayley. Assessment of residual stress prox-imal to the dent repair area of HMCS. DRDC-Atlantic Tech-nical Memorandum TM2006–275. November 2006.

7. C. Bayley and T. Romans. Comparison of Pressure Hull Straightening Approaches: Mechanical and Flame Straigh-tening. DRDC-Atlantic Technical Memorandum TM 2007– 184. October 2007.

8. T. Hutchings, P. Withers, T. Holden, and T. Lorentzen. In-troduction to the characterization of residual stress by neu-tron diffraction. CRC Press, Taylor and Francis Group. 2005. ISBN-13: 978–0-415–31000–0.

9. K. Masubuchi. Analysis of welded structures. Massachusetts Institute of Technology, Pergamon Press, New York, USA. 1981.

10. N. Pegg and M. Smith. Assessment of damage to the pres-sure hull of HMCS VICTORIA. DRDC-ATLANTIC-TM-2002–162, 01 January 2003.

11. T. Boellinghaus, T. Kannengiesser, M. Neuhaus, and W. Floring. Analysis of residual stress distribution in welded joints depending on the restraint intensity. Federal Institute for Materials Research and Testing, Germany.

12. M.F. Kanninen, F.W. Brust, J. Ahmad, and I.S. Abou-Sayed. The numerical simulation of crack growth in weld-induced residual stress fields. Battelle Columbus Laboratories, Co-lumbus, Ohio, USA. 1981.