**H G Pisarski and A Muhammed**

Paper presented at 3rd International Pipeline Technology Conference, Brugge, Belgium, 21-24 May 2000

Following ductile failure of some girth welds during installation of a 10in diameter, API 5L X60 offshore oil pipeline, subsequent testing indicated that the cleavage resistance of the weld metal was variable and might be poor at the minimum design temperature of -10°C. Although this finding had no direct bearing on the ductile failure that had occurred, the integrity of already laid girth welds for trenching operations and service was called into question. This paper describes the fracture mechanics tests and analyses, which included CTOD tests and wide plate tests, that were conducted to assess fitness-for-service of the installed sub-sea pipeline girth welds. It is shown that despite the poor toughness, the girth welds are tolerant to possible fabrication flaws and that they are fit-for-service.

## 1. Background

The installation procedure for this particular offshore pipeline involves bending the pipe through a controlled radius to give a maximum outer fibre strain of 1.7% (for a 10in diameter pipe) in order to achieve a steep departure angle from the vessel. It was at this stage that the ductile failures occurred. These were associated with buckling whilst the pipe was still warm because of residual heat remaining after welding. The causes of these failures are not the subject of this paper. However, following the failures, adjacent girth welds were removed for testing purposes. These tests showed that the weld metal had variable fracture toughness. At the minimum design temperature of -10°C, the CTOD values ranged from 0.03mm to 0.54mm. In addition, Charpy values as low as 16J at 0°C were measured. These minimum values of fracture toughness are below what would be deemed acceptable at Tier 3 of the EPRG guidelines on the assessment of defects in girth welds. ^{[1]} Consequently, the results called into question the low temperature fracture resistance of the girth welds already laid for subsequent trenching operations and service. It is these aspects of fitness-for-service that are considered in this paper.

The girth welds were made by manual metal arc welding using cellulosic electrodes (E8010 type) for the root and hot pass, and low hydrogen electrodes (E9018 type) for the fill and cap. The pipe was to API 5L X60 with an outside diameter of 273.1mm and wall thickness of 14.3mm and supplied coated with polyurethane insulation. After welding, the girth welds are inspected by radiography and then coated with thick polyurethane insulation.

## 2. Material properties

A total of 27 CTOD specimens (Bx2B, a/W = 0.5 specimens to BS 7448) from six different production girth welds were tested at -10°C and 0°C. These were notched along the weld centre line. The girth welds had experienced varying degrees of straining during installation. The nominal strain experienced by the pipe was 1.7%. Statistical analysis of the results showed that there was not a significant effect of temperature, so they were combined for subsequent analysis. Since the assessment procedure was to be K-based, the CTOD values were converted to K _{CTOD} values using the following equation that was developed in the SINTAP project. ^{[1]}

where:

σ_{f} δ E ν |
= = = = |
weld metal flow stress (average of yield and tensile strengths) CTOD Young's modulus, 207,000N/mm ^{2} Poissons ratio, 0.3 |

(It should be noted that this equation is intended to be applicable only for deeply notched bend specimens taken from steels with low strain hardening behaviour).

Statistical analysis of the K _{CTOD} values using a maximum likelihood method, involved censoring those representing maximum load ( δ _{m}) values. The log-normal distribution was found to provide a best fit to the data, as shown in *Fig.1.* The mean value was 7548N/mm ^{3/2}, and the mean minus one standard deviation was 4120N/mm ^{3/2}. According to BSI PD6493:1991 and BS 7910:1999, it is the later value that is recommended to be used in deterministic fracture mechanics calculations to assess flaw significance.

Eleven all-weld metal tensile test results were available; these gave average yield and tensile strengths of 606N/mm^{2} (standard deviation 50N/mm^{2}) and 681N/mm^{2} (standard deviation 31N/mm^{2}), respectively. These compare with average yield and tensile strengths of 488N/mm^{2} and 585N/mm^{2} (standard deviation was approximately 20N/mm^{2} for both parameters) obtained from 90 tests given in the pipe mill certificates. Thus, the nominal degree of weld metal strength overmatch was approximately 24%.

## 3. Fracture mechanics analysis

Two types of fracture mechanics analyses were conducted; deterministic and probabilistic. Both were based on the BSI PD6493:1991 Level 2 FAD flaw assessment procedure. This procedure assess the likelihood of failure with respect to fracture and local plastic collapse from a flaw, for a given geometry and set of stress conditions. The deterministic analyses were conducted using TWI software Crackwise 2, and the probabilistic analyses were performed using in-house software FORM and MONTE which are based on first order reliability and Monte Carlo simulation methods. The probabilistic analyses estimate the probability of an assessment point being outside the failure locus in the Level 2 FAD. In other words, an unacceptable result according to BSI PD6493:1991. However, this does not necessarily imply failure of the component. Further details about these analyses are given in the next sections.

### 3.1 Deterministic Analyses

A summary of the main input data for the deterministic analyses is presented in *Table 1.* Fracture toughness, K_{CTOD}, was based on the mean minus one standard deviation, as described earlier. A set of calculations was conducted assuming that tensile properties of weld metal governed local plastic collapse, and another set that plastic collapse was governed by the specified minimum yield and tensile strength of API 5L X60 pipe (a very conservative assumption). Separate stress analyses indicated that during trenching the pipe is in bending with a maximum axial outer fibre stress (at the 12 o'clock position) of 302N/mm^{2}. During service, the maximum axial stress is approximately 160N/mm^{2}, but this is estimated to reach 260N/mm^{2} if the bottom of the trench is not flat and there is rock or prop present. In addition to service stresses, allowance was made for residual stresses remaining after welding. However, these are likely to be modified as a result of plastic strains developed during installation and will vary around the circumference of the pipe. Since their magnitude was uncertain, they were assumed to be tensile and initially equal to the parent pipe yield strength, but were allowed to be relaxed by the applied stress in accordance with equations given in BSI PD6493:1991.

**Table 1: Basic input data for deterministic analysis**

Pipe diameter: | 273.1mm |

Pipe wall thicknesses: | 14.3mm |

Misalignment: | 2mm |

Primary stress: | 302N/mm^{2} trenching260N/mm ^{2} service at prop160N/mm ^{2} service no prop |

Residual stress: | Parent yield initially, 488N/mm^{2} |

Weld metal yield strength: | 606N/mm^{2} |

Weld metal tensile strength: | 681N/mm^{2} |

Fracture toughness, K _{mat}: |
4120N/mm^{3/2} |

Plastic collapse: | Kastner equation for S _{r} |

It is recognised that the girth welds may not be perfectly aligned because of axial misalignment and differences in pipe wall thickness. The combined effect of these was assumed to be equivalent to a maximum alignment of 2mm. A subsequent review indicated that the maximum misalignment would not exceed 1.5mm, consequently, the value used is conservative. The effect of misalignment is to give rise to local bending which will affect the risk of fracture and plastic collapse. The analyses employed the Kastner equation for the calculation of net section stress and the estimation of local collapse (S r axis of the FAD).

Finally, a reference flaw size was chosen. As the welds were inspected by radiography and passed as acceptable to BS 4515 criteria, the reference flaw was chosen to represent a possible fabrication flaw which might just escape detection; this was assumed to be surface breaking, not exceed one weld bead height (3mm maximum) and with the maximum length acceptable to BS 4515, viz 25mm.

The results of the calculations, in terms of maximum tolerable flaw height (for surface flaws) versus maximum tolerable flaw length, are shown in *Fig.2*. For all the cases considered, there is a significant margin of conservatism between the tolerable flaw sizes and the reference flaw. (Similar conclusions were drawn for embedded flaws, but the results are not shown here). Conservatism is reduced slightly, if the analysis is based on the parent pipe properties. However, this is due to the conservative assumption that local collapse for a weld metal flaw is governed by the flow strength of the parent pipe.

### 3.2 Probabilistic Analysis

In the probabilistic analyses the input data were similar in many aspects to the deterministic analyses except that statistical distributions were chosen to represent each input variable. As far as possible, the distributions were based on actual data and these are summarised in *Table 2*.

**Table 2: Input data for probabilistic fracture mechanics analyses**

Variable | Distribution type | Mean/Scale | Standard Dev/Shape | Source |
---|---|---|---|---|

Wall thickness, mm Bending stress due to misalignment, P _{b}/P_{m} |
Weibull Normal |
14.56* 0.135 |
26.29** 0.033 |
Measurement Based on wall thickness |

Primary axial stress, N/mm^{2} trenching service |
Normal Normal |
302 160 |
15 8 |
Assumed Assumed |

Residual stress, N/mm^{2} Case 1 Case 2 |
Normal Normal |
488 98 |
20 195 |
Parent yield Review |

Weld metal yield strength, N/mm^{2} Weld metal tensile strength, N/mm ^{2} |
Normal Normal |
606 681 |
50 32 |
Measurement Measurement |

Fracture toughness, K_{CTOD}, N/mm^{3/2} |
Log-normal | 8.9291 | 0.6054 | Measurement |

Flaw height (a), surface embedded (half height, mm) |
Log-normal Weibull |
0.7399 1.334 |
0.2341 3.806** |
UT measurement UT measurement |

Flaw height (half length c, mm) | Log-normal | 1.7511 | 0.5375 | UT measurement |

* Scale parameter for Weibull distribution ** Shape parameter for Weibull distribution |

All the input variables were considered to be independent. Thus the effect of variations in wall thickness was considered independently of the bending stress that this causes. The primary axial stress was assumed to be subject to a coefficient of variation of 5%. Fracture toughness was taken from the log-normal parameters established from K_{CTOD} described earlier ( *see Fig.1*). With respect to residual stresses, two calculations were conducted. The first made exactly the same assumptions as the deterministic analyses, except that the parent pipe yield strength was not a fixed value. The second was based on a review of transverse residual stresses in pipe girth welds by Mohr. ^{[4]} This indicated that for the pipe size and heat input employed to make the girth welds, the residual stresses at the weld root of single sided welds are scattered and could vary between 0.6 _{σ YP} (compressive) and _{σ YP} (tensile). Other work by TWI on plastically strained pipe indicated residual stresses to be between 0.06 _{σ YP} and 0.3 _{σ YP} . For the purposes of these analyses, residual stresses were assumed to be normally distributed with a mean of 0.2 _{σ YP} and standard deviation of 0.4 _{σ YP} (where _{σ YP} = 488N/mm ^{2}).

Finally, flaw size distributions (height and length) were derived from an analysis of automated ultrasonic inspection data obtained from another pipeline project which employed similar consumables and welding procedures. Both surface breaking and embedded flaws were considered; the latter was assumed to be located at a fixed distance below the surface of 3mm.

For surface flaws, the mean flaw height was 2.1mm and the mean length was 11.5mm. The distribution employed is considered to be conservative as it excludes the effect of repair which removes the larger flaws.

The results of the analyses are summarised in *Table 3* and give the probability of failure per flaw (P_{1}). In order to establish the probability of failure for the whole pipeline (P_{f}), the probability of having a flaw in the maximum tension zone of the weld (P_{2}), the probability of a flaw in a weld (P_{3}) and the number of welds to be considered (N) were substituted in the approximate relation:

P_{f} = P_{1} . P . P_{3} . N [2]

P_{2} can be assumed to be approximately 0.1 (maximum tension zone as proportion of circumference). P_{3} depends on the number of repairs. A typical repair rate for an offshore pipeline is 0.5% to 1%. For a pipeline that contains 1000 girth welds this means that approximately 5 to 10 welds will be repaired. Assuming 80% of 'unacceptable' flaws in the repaired welds are removed, means that between 1 and 2 welds might contain flaws, thus P_{3} is approximately 10^{-3}. Thus the P_{1} figures given in *Table 3* can be reduced by a factor of 10 to provide an estimate of the probability of failure for the pipeline.

**Table 3: Results from probabilistic fracture mechanics analyses**

Stress case | Flaw type | Residual stressassumption | Probability offailure/flaw, P _{1} |
---|---|---|---|

Trenching | Surface | _{σ YP} (Case 1)0.2 _{σ YP} (Case 2) |
2E-2 6E-3 |

Trenching | Embedded | _{σ YP} (Case 1)0.2 _{σ YP} (Case 2) |
7E-3 2E-3 |

Service | Surface | _{σ YP} (Case 1)0.2 _{σ YP} (Case 2) |
1.3E-2 2E-3 |

Service | Embedded | _{σ YP} (Case 1)0.2 _{σ YP} (Case 2) |
3E-3 4E-4 |

However, the probability of failure will be lower than calculated, because the analyses made no allowance for the differences in constraint between the girth weld containing a flaw (low constraint) and the highly constrained deeply notched bend specimen from which the fracture toughness data were derived. To some extent, differences in constraint can be allowed for by incorporating modelling uncertainty in the assessment line of the Level 2 FAD. Recently, this has been explored by Muhammed et al [5]. They show that where failure is likely to be elastic-fracture dominated, the inclusion of modelling uncertainty in the probabilistic analysis implies approximately one order of magnitude lower probability of failure. This means that the failure probabilities given in *Table 3* may be reduced by two orders of magnitude overall, to obtain estimates for the pipeline. As a consequence, the probabilities of failure estimated for these girth welds fall within the target ranges (10^{-2} to 10^{-5}) considered acceptable by Sotberg et al ^{[6]} as established in the SUPERB offshore pipeline project. Trenching falls within the ultimate limit state 'low' safety class, where failure would imply negligible risk to human safety and the environment. The lower probabilities of failure estimated for the service condition are within the 'normal' to 'high' safety class target failure probabilities (10^{-3} to 10^{-5}).

## 4. Wide plate tests

As a back-up to the fracture mechanics tests, deterministic and probabilistic analyses, two wide plate tests were conducted to simulate axial loading of the girth weld when it contains a surface breaking weld root crack. It was not possible to test actual production girth welds, as the pipeline had already been laid (but not trenched). Consequently, welds simulating the girth welds were made using the same consumables and parent pipe. To facilitate testing, the pipe was cut longitudinally and flattened prior to welding. The resulting wide plate specimen was 350mm wide with a gauge length of 380mm; the weld was transverse to the loading direction, see *Fig.3*. However, CTOD fracture toughness tests conducted on these welds (i.e. extensions to the wide plate specimens) indicated that their toughness was higher than the production girth welds. Similar toughness (in terms of scatter and minimum values) could be achieved by testing at -30°C, see *Fig.4*. Consequently, the wide plate tests were conducted at -30°C (20°C below the actual minimum design temperature for the pipeline). The wide plate specimens containing a semi-elliptical surface cracks in the weld metal representative of a lack of weld root penetration flaw, nominally 3mm high and 50mm long. In addition, the wide plates contained intentional axial misalignment of approximately 1mm in the first specimen and 0.6mm in the second.

At -30°C, neither specimen fractured. The first wide plate specimen survived a strain of 4.99%, gross stress of 653N/mm^{2} and attained a CTOD of 1.17mm. The second survived a strain of 2.04%, a gross stress of 596N/mm^{2} and attained a CTOD of 1.13mm.

Post test metallography showed that in the first wide plate specimen the maximum crack depth was 2.42mm but that it had extended by tearing to a depth of 3.04mm. The crack tip was located in the cellulosic weld metal at the weld root. In the second wide plate specimen, the initial crack depth was 4mm and it had extended by tearing by 0.46mm. The crack tip was located in columnar weld metal of the low hydrogen weld, this is the same microstructure which initiated the low CTOD values in the small scale bend tests.

The relationship between critical CTOD and applied stress predicted from the Level 2 FAD and CTOD measured experimentally in the wide plate specimens is shown in *Fig.5 and 6*. If the analyses ignore residual stress, then potentially unsafe predictions are possible at low applied stress. However, if yield magnitude residual stress are assumed to be present initially, conservative predictions result, but the degree of conservatism can be high. The deterministic analyses described earlier were based upon this approach. Assuming the Level 2 FAD model to be reasonably correct, then low, but not negligible, residual stresses must be present in the wide plate specimen and should be allowed for in the analyses of tolerable flaw size.

### 5. Discussion

The deterministic analyses based on BSI PD6493:1991 Level 2 FAD procedure show that there is a respectable margin of safety between the estimated tolerable flaw sizes for the trenching and operating conditions, and the reference flaw. The reference flaw is considered to represent the largest that could escape detection. This conclusion is supported by the probabilistic analyses which show that the estimated probability of failure for the pipeline is 10 ^{-3} or less, and achieves the target reliabilities recommended in the SUPERB project. ^{[5]} The wide plate tests lend further support to the analyses conducted. Neither of the two wide plate specimens failed at -30°C (20°C below the minimum design temperature) when realistic weld root flaws, approximately 3x50mm, were present. Thus on the basis of these tests and analyses, it can be concluded that the girth welds are fit-for-service, with respect to avoidance of fracture and plastic collapse, for both trenching operations and service.

One intriguing problem is why the fracture toughness tests taken from the flat plates gave lower results than the production girth welds, see *Fig.4*. An explanation for this lies in the role played by strength mismatch between the weld metal and parent pipe and the straining experienced during pipe installation. Although the weld metal overmatched the yield strength of the parent pipe by about 24%, situations arose where overmatching was not maintained after yield, see *Fig.7*. Normally this is not significant since design stresses are below SMYS. However, pipe laying involved plastic straining of the pipe to 1.7%. In practice, this nominal strain would be elevated by misalignment effects and, more importantly, by differences in stiffness between the factory applied pipe coating and field coating applied to the girth weld after welding and inspection. (Work by Crome ^{[7]} has shown that differences in coating stiffness can induce buckling failures). Subsequent analyses indicated that the effect of these would be to locally increase the strain to about 3.9%. This means straining of the weld metal would take place with a consequent reduction in fracture toughness. This reduction in fracture toughness is illustrated by the results of fracture toughness tests conducted on the wide plate specimens before and after testing, see *Fig.8*. 2% strain applied to the wide plate specimen produces no reduction in weld metal toughness, since only a small amount of plastic straining can occur in the weld metal. 5% strain in the wide plate causes plastic straining of the weld metal and the toughness is reduced significantly.

## 6. Conclusions

Despite the occurrence of scatter and low CTOD fracture toughness values in production girth welds from an oil pipeline, analyses of the results by fracture mechanics procedures using both deterministic and probabilistic methods have shown that they are fit-for-service. These analyses were backed up by wide plate tests which showed that the welds could tolerate at least up to 5% strain without failure with a 3x50mm flaw present in the weld root.

Analyses showed that the low toughness measured in the production girth welds was partly caused by plastic straining of the weld metal during installation, despite the weld metal overmatching the yield strength of the parent pipe.

## References

- G. Kauf and P Hopkins. The EPRG guidelines on the assessment of defects in transmission pipeline girth welds. 3R International, 35, 10/11, 620-624, 1996.
- SINTAP. Structural integrity assessment procedures for European industry. Final procedure, November 1999. European Union Brite-Euram Programme: BE95-1426, BRPR-CT95-0024).
- BSI PD6493:1991. Guidance on methods for assessing the acceptability of flaws in fusion welded structures. British Standards Institute, London. Now replaced by BS 7910:1999.
- W.C. Mohr. Internal surface residual stresses in girth butt-welded steel pipes. PVP-Vol.327. Residual Stresses in Design, Fabrication, Assessment and Repair ASME 1996, 37-45.
- A. Muhammed, H.G Pisarski and A. Stacey. Using wide plate test results to improve predictions from probabilistic fracture mechanics. 13
^{th}European Conference on Fracture, ECF 13, 6-9 September 2000, San Sebastian, Spain (to be published). - T. Sotberg, I. Moon, R. Bruschi, G. Jiao and K.J. Mork. The SUPERB project: recommended target safety levels for limit state based design of offshore pipelines. 1997 OMAE-Vol.V, Pipeline Technology, ASME1997, 71-77.
- T. Crome. Reeling of pipelines with thick insulating coating, finite element analysis of local buckling. Offshore Technology Conference, Houston, Texas, May 1999, Paper OTC 10715.