Pedro M. Vargas
Chevron ETC
Houston, Texas, USA
Stig Wästberg
Det Norske Veritas (DNV)
Høvik, Norway
Paul Woollin
TWI
Granta Park, Cambridge UK
Paper presented at 28th International Conference on Offshore Mechanics and Arctic Engineering (OMAE 2009), Honolulu, Hawaii, 31 May  5 June 2009.
Abstract
Following the failure of several subsea components made of duplex steel, two JIPs were formed, one by TWI and another by DNV and Sintef to address the failure mechanism and to formulate design guidance for the industry. (TWI: The effects of notches and welds on hydrogen embrittlement stress cracking of duplex stainless steels, Sintef/DNV: HISC) Hydrogen charging from the cathodic protection system in the presence of creep strains embrittles the duplex steel, making the duplex susceptible to cracking (hydrogeninducedstresscracking, HISC). Creep effects focused on strain measurements in the test specimens from early work at TWI, favoring a strain based approach in the development of early versions of the design guidance for the industry. This paper summarizes the relevant content from the two JIPs to formulate a stress based design criteria, and provides new FEA assessment of the Foinhaven Hubs to better quantify the effect of residual stresses. The basis for the stressbased design guidelines in DNVRPF112 is presented that promises to be easier to apply and equally robust as the strainbased approach.
Introduction
Duplex subsea equipment failures to date have been studied extensively and they can be categorized into two types: 1) Failures in large forgings due to excessive loading, at times combined with unfavorable coarse austenitic spacing, and 2) fillet/socket welds that are underdesigned and/or with high ferritic content.
For the forgings, the failure did not typically occur at the weld toe, but instead at a stress concentration location removed but in proximity of a girth weld. The fracture surface showed signs of hydrogen embrittlement in the form of transgranular brittle features. The areas of failure were not coated.
For the fillet/socket welds, due to undersized welds or high loading, overloading was found to be a culprit, and coatings were either not present or damaged.
These HISC failures have initiated from the outer surface which is exposed to cathodic protection.
The experience base suggests that there have been no failures where all the requirements set by the design codes have been fulfilled for the material, fabrication or loading at all times before HISC occured. The reported failures have been in fillet welds not in accordance with code (high ferrite content and/or lack of fusion) and external forces that exceeded design loads.
The level of reliability of these duplex components has been shown to be little less than what industry expects, i.e., most designs can tolerate a little overload/bad design, and still function adequately. Conservatisms built into the codes generally help make this happen. DNVRPF112 focuses on providing limits against stress and strain that directly address HISC.
Nomenclature
σ_{m} : 
through wall membrane stress. 
σ_{b} : 
through wall bending stress. 
σ_{m+b} : 
through wall membrane+bending stress. 
R : 
outside radius 
t : 
wall thickness 
γ_{HISC} : 
material factor for coarse grain duplex material 
σ_{yield} : 
0.2% offset yield stress, also called proof stress 
SMYS : 
Specified minimum yield stress, also σ _{yield} 
L_{res} : 
Critical distance for residual stress effects, 2.5√Rt. 
HISC mechanism
Although the interaction between hydrogen, microstructure, stress and strain is not well understood, phenomenologically HISC has been found to occur when cathodic hydrogen charging occurs while the duplex matrix undergoes creep. It is postulated that as the duplex matrix undergoes creep, hydrogen diffuses through the ferrite, recombining and locking itself into vacancies produced, embrittling the ferrite phase. Cracks occur in the ferrite phase, and link up across austenitic phases as the creep progresses (Figure 1).
The HISC mechanism, although local due to the surface hydrogen concentration, requires the potential for creep strains to occur in the presence of hydrogen diffusion. For smooth axially loaded specimens, since the entire crosssection is at the same nominal stress state, the local stress is equal to the nominal stress state. Then the amplitude of this stress state gives the potential of the entire crosssection for creep, including the local surface with hydrogen, making it a valid parameter for the assessment of HISC. For a more complex stress state (a notched bar for example), however, the local stresses may be high but if the nominal crosssection stresses are low enough that sufficient accumulation of creep strains at the surface is not possible, then HISC will not occur.
Therefore the stress based HISC avoidance criteria focuses on throughwall structural stresses such as that used in ASME codes for pressure vessels: σ_{m} for the throughwall membrane stress and, σ_{m+b} for the throughwall membrane + bending stresses. These structural stresses give more information on the gross creep potential of the wall than the local peak stresses. This has been validated through a significant amount of testing using notched round bars, single edge notched tensile specimens and full scale girth welds.
Figure 2 shows the separation of these stresses from an arbitrary stress gradient. Since the stress based criteria addresses the notch acuity effects separately, and conventional design practices limit any significant effect on the throughwall structural stresses, linear elastic finite element analyses is sufficient to obtain the stresses needed.
Fig.2. Membrane and bending stresses
Stress criteria for HISC avoidance
The stress criteria focuses on providing limits to the throughwall membrane and membrane+bending stresses to avoid loadings that can impose gross membranestretching creep or gross throughwall bending creep. Then a penalty is placed: 1) on the presence of stress risers, and 2) on locations that are in the vicinity of girth welds. In addition a material quality factor is applied that reduces the allowable limits for coarse austenite spacing. All through wall sections for a subsea duplex component must satisfy both stress checks:
Membrane stress check:
σ_{m} <80%σ _{yield} 
all throughwall sections, everywhere. (TWI, HISC1, HISC2) 
Membrane+bending stress check: 
σ_{m+b} <100%σ _{yield} 
smooth sections without stress raisers or welds, outside of L_{res}, (HISC1, HISC2) 
σ_{m+b} <90%σ _{yield} 
smooth sections within L_{res} = 2.5√Rt of a girth weld (Foinhaven Hub FEA) 
σ_{m+b} <90%σ _{yield} 
Weld toes, attachments, and stress risers outside of L_{res} = 2.5√Rt of a girth weld (HISC2) 
σ_{m+b} <80%σ _{yield} 
Weld toes, attachments, and stress risers within L_{res} = 2.5√Rt of a girth weld (Foinhaven Hub FEA) 
Material Penalty: 
For materials with coarse grain austenite spacing, the stress limits above are multiplied by γ _{HISC} = 0.85. (TWI, HISC2, Foinhaven Hub FEA) 
The italics in the parenthesis highlight the JIP data that supports the limit. The Foinhaven FEA is presented in a later section.
The DNV RPF112 also requires the same checks with the vonMises (equivalent stresses) for these stresses and provides guidance.
TWI JIP summary
TWI performed a comprehensive study of the HISC performance of smooth tensile coupons (3.75mm diameter) with 18 different materials that included forgings, bars plates and pipes. Over 200 specimens were tested. Different austenitic spacing, weld materials and simulated HAZ were also part of this matrix. Each sample was kept under tension in a small polymer vessel containing natural seawater and polarized by potentiostat. Different temperatures, charging potentials, hydrostatic pressures and test durations were also examined. The test results found important trends with respect to cathodic charging levels and austenitic spacing, but also found that proximity of the nominal stress state to yield was the most prominent indicator of the propensity for HISC. The JIP indicated that:
 Regardless of material, as long as the applied nominal stress was less than 87% of 0.2% offset yield, HISC did not occur.
 Regardless of material, as long as the applied initial nominal strain was less than 0.5%, HISC did not occur.
These two observations held for all of the materials tested: coarse and fine austenite spacing, welds and the HAZ simulated specimens. The study also found that prestraining up to 3% provided resistance to creep and HISC. Figure 3 shows some selected results for materials tested at TWI.
Subsequent extension of the JIP focused on the notch acuity and full scale performance of girth welds. The results from this study indicated that the notch acuity did not affect the HISC performance of the test specimens. A full scale girth weld test also showed that membrane stresses in the pipe in excess of yield was needed for HISC to occur.
These results form the basis for:
 The 80%σ_{yield} limit on the average membrane stresses.
 The use of structural stresses instead of notch stresses for HISC assessment.
 The need for a material degradation factor to account for austenite spacing.
 Localselfequilibrated residual stresses have little or no effect on HISC resistance.
Fig.3. HISC stresses for Garn West and Foinhaven
HISC1 and HISC2
HISC1 and HISC2 did a significant amount of testing using three specimen geometries: 1) Smooth, 2) Single edge notch with a Vnotch, and 3) Single edge notch with a Unotch. These specimens were tested under constant load. The Vnotch specimen was designed to incur the notch stress concentration equivalent to a typical weld toe in the duplex structure. Due to the onesided notch, a cross sectional moment is created in the specimen even though loaded axially.
Figure 4 shows the specimens and Figure 5 shows the notch dimensions and the 2D finite element meshes used to compute the stresses and strains in the specimens. The specimens are 12mm height by 9mm in thickness.
Fig.4. SENT specimens from HISC
Fig.5. SENT specimens from HISC
The loading frame provides significant flexibility, and for the small deformations leading to HISC cracking, pinended boundary conditions were used in the analyses. Using the creep constants derived in HISC, a strain hardening creep law implemented in ABAQUS^{[7]} was used. The stresses across the Unotch and Vnotch specimens are shown in Figures 6 and 7. Note that a significant amount of bending is present.
Fig.6. Longitudinal stresses in Unotch specimen @ 80% proof load
These figures show the stress distribution immediately after load application, and the stresses after one month worth of creep has occurred. Note that the peak stresses drop significantly due to creep and that due to plasticity and creep, the stress at the notch is not the maximum stress; the maximum longitudinal stress occurs at a small distance into the specimen. Even though there is significant change in the peak stress due to creep, most of the cross section is at stresses less than 60% of the proof stress both immediately after load application and after 1 month worth of creep. Locally the creep strains relax the peak stresses but overall the elastic stress state of over 90% of the cross section will restrain deformation and therefore accumulation of additional creep.
Fig.7. Longitudinal stresses in Vnotch specimen @ 80% proof load
Table 1 summarizes the computed results. The 0.2% offset yield stress was used as the proof stress, 600 MPa. Four columns are reported for each specimen: 80%, 90%, 100%, and 110% proof load average membrane loading. HISC finite element studies report the elastic SCF, defined as the ratio of the peak stress at the notch to the average membrane stress across the cross section at the notch, to be 6.4 for the Vnotch and 3.2 for the Unotch. Figures 6 and 7 attest to the futility in using elastic stress concentration factors in the quantification of HISC; the potential for creep is not reflected in the local stress state at the notch.
The structural stress definition of SCF = σ_{m+b}/σ_{m}, ratio of the membrane+bending stresses to the applied nominal membrane stress at the notch cross section, is also reported in Table 1. The average membrane stress is in the plane of the notch, not the nominal remote stresses in the unnotched body of the specimen. These confirm that the SENT specimens are undergoing significant bending: approximately 40% for the V=notch and 50% for the Unotch specimens. The Unotch specimens have more bending due to the larger notch; the applied eccentricity of the loading is higher.
Table 1: Stresses in HISC Specimens
Unotch  80% Proof  90% Proof  100% Proof  110% Proof 
(MPa)  % Proof  (MPa)  % Proof  (MPa)  % Proof  (MPa)  % Proof 
Load 
m 
480.122 
80.0% 
540.288 
90.0% 
603.32 
100.6% 
661.028 
110.2% 
b 
276.158 
46.0% 
308.169 
51.4% 
334.731 
55.8% 
358.528 
59.8% 
m+b 
756.28 
126.0% 
848.457 
141.4% 
938.051 
156.3% 
1019.56 
169.9% 
p 
2.51204 
0.4% 
74.9184 
12.5% 
144.919 
24.2% 
198.579 
33.1% 

SCF 
1.58 
1.57 
1.55 
1.54 
Creep 
m 
480.293 
80.0% 
540.762 
90.1% 
601.482 
100.2% 
665.999 
111.0% 
b 
271.498 
45.2% 
294.2 
49.0% 
306.393 
51.1% 
306.202 
51.0% 
m+b 
751.792 
125.3% 
834.962 
139.2% 
907.875 
151.3% 
972.201 
162.0% 
p 
141.719 
23.6% 
192.16 
32.0% 
226.674 
37.8% 
232.687 
38.8% 

SCF 
1.57 
1.54 
1.51 
1.46 
Vnotch  80% Proof  90% Proof  100% Proof  110% Proof 
(MPa)  % Proof  (MPa)  % Proof  (MPa)  % Proof  (MPa)  % Proof 
Load 
m 
480.269 
80.0% 
540.463 
90.1% 
600.675 
100.1% 
661.077 
110.2% 
b 
200 
33.3% 
223.963 
37.3% 
246.427 
41.1% 
264.654 
44.1% 
m+b 
680.269 
113.4% 
764.426 
127.4% 
847.102 
141.2% 
925.731 
154.3% 
p 
169.608 
28.3% 
108.537 
18.1% 
55.2359 
9.2% 
23.3327 
3.9% 

SCF 
1.42 
1.41 
1.41 
1.40 
Creep 
m 
480.435 
80.1% 
540.788 
90.1% 
599.755 
100.0% 
665.014 
110.8% 
b 
197.737 
33.0% 
216.429 
36.1% 
227.365 
37.9% 
236.833 
39.5% 
m+b 
678.172 
113.0% 
757.217 
126.2% 
827.12 
137.9% 
901.847 
150.3% 
p 
16.7974 
2.8% 
57.496 
9.6% 
55.0239 
9.2% 
91.9798 
15.3% 

SCF 
1.41 
1.40 
1.38 
1.36 
m: σ_{m}, average membrane stress
b: σ_{b}, linearized bending stress
m+b: σ_{m+b} : σ_{m} + σ_{b}
p: σ_{p}, peak stress such that σ_{m} + σ_{b} + σ_{p} = σ_{notch}
Figure 8 shows the normalized net section stress (again at the notch cross section) as a function of time to HISC failure in hours. Note that the net section average membrane stress limit of 80% is validated, although one Vnotch specimen fails at 78% average membrane loading of the proof load.
Fig.8. Normalized net stress vs. time to HISC failure
Figure 9 shows the normalized net section membrane+bending stress (again at the notch cross section) as a function of time to HISC failure in hours. The normalized net section stress is multiplied by 1.4 for the Vnotch and 1.5 for the Unotch specimens. This forms the basis for the structural stress limit proposed for HISC assessment. For the membrane+bending stresses, the HISC limit is significantly higher. A membrane+bending stress limit of 100% of the proof load is validated in this plot. Note that the single violation point is that of a smooth specimen (i.e. no bending), and would violate the 80% membrane criteria.
Fig.9. Normalized net membrane+bending stress vs. time to HISC failure
Figure 8 also shows that approximately 10% of proof loading separates the Unotch from the Vnotch when comparing the structural stress state in the HISC specimens. This forms the basis for accounting for the notch effects at the weld toes of duplex materials.
Other relevant observations regarding residual stresses and prestraining:
 Residual stresses were introduced into a Vnotch specimen through plastic deformation. The resulting specimen showed no difference in HISC resistance than a virgin specimen (HISC2).
 Prestraining was found to mitigate HISC, the streses needed to cause HISC increased with the strainhardening (TWI).
 Coarse austenite spacing was found to decrease HISC resistance by approximately 15% (TWI/HISC2).
These results form the basis for:
 The 80%σ_{yield} limit on the structural membrane stresses.
 The 100%σ_{yield} limit on the structural membrane+bending stresses.
 The use of structural stresses instead of notch stresses for HISC assessment.
 A 10% proof load penalty due to notch effects at weld toes.
 Localselfequilibrated residual stresses have little or no effect on HISC resistance.
 The material degradation factor to account for austenite spacing, γ_{HISC}=0.85.
Foinhaven hub full scale tests
TWI performed full scale testing of the Foinhaven Duplex hubs. These tests proved to be controversial in that a significant penalty was implied to account for residual stresses in the initial strain based HISC assessment. Figure 10 shows the testing arrangement. The duplex material had a coarse grain structure with austenitic spacing measuring 53µm transverse to the grain orientation and 128µm along the grain orientation. The proof loading is reported as 570 MPa. The strain hardening creep model from the HISC2 JIP was modified to give similar creep behavior at the same applied stresstoyield ratio (Figure 11).
Residual stress measurements showed that the outside was in tension both axially and circumferentially, while the inside of the nib and tube was in compression. Axial residual stresses on the outside surface were in the range of 368468 MPa on the nib and 112 to 369 MPA on the tube. Axial stresses measured on the inside surface were in the range of 264 to 546 MPa.
Fig.10. Foinhaven hub testing at TWI
Fig.11. Foinhaven hub strain hardening creep model
Two full scale tests of the Foinhaven hubs were performed identified as P16 and P15. Cathodic polarization to 1100mV was applied in seawater, with a one week precharging prior to load application. P16 was tested first, and 13 increments of increased step loading were applied to finally fail at critical strain of 0.24%. Bending on the nib and tube was applied through tightening of the nut on the tension bar in Figure 10. Strain was monitored at gage#10 in P16, and gage#1 in P15, and then related to the critical cracking location through calibration loadings and validated through FEA analyses. The critical strain was found to be 2.03 times the gage strain. P15 failed after 2 increments of increased load at a critical strain of 0.22%.
Both failed at similar load levels. Using 20noded bricks, the hub was modeled as shown in Figure 12. Strain monitoring gage locations and the critical strain location is shown. Proper boundary conditions were applied to correctly apply the tension bolt loading. The multistep loading was simplified to a single step. Three cases were run:
 Residual stresses are assumed to be negligible. The bolt tensile loading is applied, and then creep is allowed for one month.
 Residual stresses are approximated through shrinkage of 0.003 strain in the weld. The bolt tensile loading is applied, and then creep is allowed for one month.
 Upper bound residual stresses are approximated through shrinkage of 0.0045 strain in the weld. The bolt tensile loading is applied, and then creep is allowed for one month.
Fig.12. Foinhaven hub FEA model
The contour plot in Figure 13 shows the longitudinal stresses at load application for Case 1. In case 1, the residual stresses of the weld is neglected, and only the applied loading via the bolt is applied. Note that the normalized longitudinal stresses approach 0.9 at the critical location on the surface. However, since most of the cross section remains elastic no creep occurs, and the stress state remains the same after 1 month of creep.
Fig.13. Foinhaven hub case 1 longitudinal stresses: no residual stresses
The contour plot in Figure 14 shows the longitudinal stresses at the end of the one month worth of creep for Case 2. The blue curve shows the residual stresses due to the 'rubberband' effect that the weld shrinkage (0.003 strain) has upon cooling. Note that the normalized longitudinal stresses exceed 1.2 at the critical location on the surface once the load is applied. In this case, the residual stress together with the applied stress is sufficient to cause creep near the surface where the hydrogen charging occurs. After 1 month worth of creep, the normalized longitudinal stress drop to 1.0 at the critical location on the surface.
Fig.14. Foinhaven hub case 2 longitudinal stresses: residual stresses due to 0.003 shrinkage
The contour plot in Figure 15 shows the longitudinal stresses at the end of the one month worth of creep for Case 3. The blue curve shows the residual stresses due to the 'rubberband' effect that the weld shrinkage (0.0045 strain) has upon cooling. Note that the normalized longitudinal stresses approach 1.3 at the critical location on the surface once the load is applied. As in Case 2, the residual stress together with the applied stress is sufficient to cause creep near the surface where the hydrogen charging occurs. After 1 month worth of creep, the normalized longitudinal stress drop to 1.0, identical to Case 2, at the critical location on the surface. The stress distribution for Case 2 and Case 3 are nearly identical after creep has taken place.
Table 2 summarizes the computed results. The 0.2% offset yield stress was used as the proof stress, 570 MPa. The three columns are for each case analyzed. The three rows cover the loading conditions imposed, residual stresses, bending load, and then creep.
The average membrane stress is in the plane of the critical section, shown in Figures 1315 as the stress path. As expected, the residual stresses are nearly all in throughthickness bending. Case 1 also serves to illustrate the computations that would results for a linearelastic FEA with out residual stresses. Ideally, a linearelastic analyses (without the need for creep assessment) would be sufficient for the HISC assessment. Since our focus is on providing a limit to the structural stresses with the objective of limiting creep deformation, the numbers of interest is in column 1 (Case 1) of Table 2. Thus, for Foinhaven full scale tests to be properly reflected in the stressbased criteria, the membrane+bending stresses will have to be limited to 74.3% of the proof loading.
Fig.15. Foinhaven hub case 3 longitudinal stresses: residual stresses due to 0.0045 shrinkage
The effect of the weld residual stress is the rubberbandlike squeezing of the tube and nib, causing tension on the outside surface and compression on the inside. This stress decays with distance. A conservative length parameter for curved shell local bending is 2.5√Rt. This residual stress effect is accounted for through a 10% penalty on locations that are within 2.5√Rt of a girth weld.
Table 2: Stresses in Foinhaven FEA
 Case 1  Case 2  Case 3 
(MPa)  % Proof  (MPa)  % Proof  (MPa)  % Proof 
Residual 
m 
0 
0.0% 
4.49568 
0.8% 
6.07091 
1.1% 
b 
0 
0.0% 
126.986 
22.3% 
170.041 
29.8% 
m+b 
0 
0.0% 
122.49 
21.5% 
163.97 
28.8% 
p 
0 
0.0% 
44.2328 
7.8% 
57.6584 
10.1% 
Load 
m 
186.169 
32.7% 
181.025 
31.8% 
177.734 
31.2% 
b 
237.145 
41.6% 
357.894 
62.8% 
392.397 
68.8% 
m+b 
423.313 
74.3% 
538.919 
94.5% 
570.132 
100.0% 
p 
109.601 
19.2% 
151.733 
26.6% 
161.77 
28.4% 
Creep 
m 
186.083 
32.6% 
177.221 
31.1% 
173.02 
30.4% 
b 
236.57 
41.5% 
335.574 
58.9% 
354.828 
62.3% 
m+b 
422.653 
74.1% 
512.795 
90.0% 
527.848 
92.6% 
p 
101.921 
17.9% 
58.4945 
10.3% 
42.4237 
7.4% 
m: σ_{m}, average membrane stress
b: σ_{b}, linearized bending stress
m+b: σ_{m+b} : σ_{m} + σ_{b}
p: σ_{p}, peak stress such that σ_{m} + σ_{b} + σ_{p} = σ_{notch}
The critical location for the Foinhaven hubs is a stress riser (14mm from weld edge) within 2.5√Rt of a girth weld (R=84.25mm, t=11.25mm, 2.5√Rt=77.0mm). The structural stress state from the linearelastic FEA would be (Case 1):
 σ_{m} = 32.7%σ_{yield}
 σ_{m+b} = 74.3%σ_{yield}
The stress limits are:
 σ_{m} <80% γ_{HISC}σ_{yield} = 68%_{HISC}σ_{yield}
 σ_{m+b} <80% γ_{HISC}σ_{yield} = 68%_{HISC}σ_{yield}
γ_{HISC}=0.85 due to the large austenite spacing, and the σ_{m+b} <80%σ_{yield} is chosen for the stress riser within 2.5√Rt of a girth weld. The stress criteria fits the failure data for the TWI Foinhaven full scale tests.
Summary
The stress criteria focuses on providing limits to the throughwall membrane and membrane+bending stresses to avoid loadings that can impose gross membranestretching creep or gross throughwall bending creep. Then a penalty is placed: 1) on the presence of stress risers, and 2) on locations that are in the vicinity of girth welds. In addition a material quality factor is applied that reduces the allowable limits for coarse austenite spacing. Figure 16 summarizes the stressbased criteria.
Fig.16. HISC stress criteria summary
References
 Stig Wästberg, Morten Solnørdal, Gustav Heiberg, Tikard Törnqvist and Pedro Vargas, Hydrogen Induced Stess Cracking, (HISC) in duplex stainless steels  DNVRPF112, Design of duplex stainless steel subsea equipment exposed to cathodic protection, OMAE200979655.
 DNV recommended practice, DNVRPF112, Design of duplex stainless steel subsea equipment exposed to cathodic protection, October 2008.
 T. S. Taylor, T. Pendlington, and R. Bird, Foinhaven super duplex BP materials cracking investigation, OTC May 1999.
 P. Woollin and A. Gregori, Avoiding hydrogen embrittlement stress cracking of ferritic austenetic stainless steels under cathodic protection, OMAE200451203
 FEM analyses of notched tension and bend specimens used in the HISC II and the Ormen Lange HISC projects, DNV technical report no. 20063259, revision no. 02, January 10, 2007.
 Roy Johnsen, Andre Mikkelsen, Bård Nyhus, Stig Wästberg, Hydrogen Induced Stress Cracking of stainless steels final report for HISC2. SINTEF report STFMKF07029, 20070926.
 ABAQUS/Standard User's Manual, Volume III, Version 6.4, Hibbitt, Karlsson & Sorensen, Inc., Pawtucket, RI 12860, www.abaqus.com.