Burst strength behaviour of an aging subsea gas pipeline elbow in different external and internal corrosion-damaged positions

Transcription

1 csnak, 2015 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~451 pissn: , eissn: Burst strength behaviour o an aging subsea gas pipeline elbow in dierent external and internal corrosion-damaged positions Geon Ho Lee, Hassan Pouraria, Jung Kwan Seo and Jeom Kee Paik The Korea Ship and Oshore Research Institute (The Lloyd s Register Foundation Research Centre o Excellence), Pusan National University, Busan Korea Received 17 March 2014; Revised 18 August 2014; Accepted 15 January 2015 ABSTRACT: Evaluation o the perormance o aging structures is essential in the oil and gas industry, where the inaccurate prediction o structural perormance can have signiicantly hazardous consequences. The eects o structure ailure due to the signiicant reduction in wall thickness, which determines the burst strength, make it very complicated or pipeline operators to maintain pipeline serviceability. In other words, the serviceability o gas pipelines and elbows needs to be predicted and assessed to ensure that the burst or collapse strength capacities o the structures remain less than the maximum allowable operation pressure. In this study, several positions o the corrosion in a subsea elbow made o API X42 steel were evaluated using both design ormulas and numerical analysis. The most hazardous corrosion position o the aging elbow was then determined to assess its serviceability. The results o this study are applicable to the operational and elbow serviceability needs o subsea pipelines and can help predict more accurate replacement or repair times. KEY WORDS: Elbow; Corrosion; Burst pressure; API X42 steel; Elbow damage; Corrosion ailure mode. NOMENCLATURE C D d d G L LF M P P O P L Curve it coeicient Speciied outside diameter o the pipe Depth o corroded region Deect actor (Goodall ormula) Deect coeicient Length o corroded region Lorenz actor Bulging stress magniication actor Failure pressure o the corroded pipe Plastic limit pressure o elbow without deect Plastic limit pressure o elbow with deect Q Length correction actor R b Elbow bend radius R m Elbow mean radius SMTS Speciied minimum tensile strength SMYS Speciied minimum yield strength t Elbow wall thickness α Circumerential angle rom the crown o the elbow γ Axial hal-angle o local thinned area γ d Partial saety actor or corrosion depth Partial saety actor or longitudinal corrosion γ m Corresponding author: Jung Kwan Seo, seojk@pusan.ac.kr This is an Open-Access article distributed under the terms o the Creative Commons Attribution Non-Commercial License ( which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

2 436 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~451 θ σ Circumerential hal-angle o local thinned area Flow stress (Goodall ormula) σ u σ y Material ultimate tensile stress Material yield stress INTRODUCTION Corrosion is a critical problem in the gas pipeline industry and elbows are one o the most corrosion prone structures in gas pipelines. It is especially important to maintain gradually corroding metal pipelines in the subsea industry. As corrosion grows, it causes material degradation in the corroded area, which inally ends in structural ailure or a burst pipe. Some studies have attempted to predict pipeline ailure in terms o the remaining strength capacity using deterministic or probabilistic approaches. Previous studies have assessed the importance o corrosion damage evaluation or numerous structures, including gas pipelines and oshore structures, and assessed their mathematical models (Bai and Bai, 2005; Bai and Bai, 2014; Kim et al., 2013; Kyriakides and Corona, 2007; Mohd et al., 2014). These techniques have been widely used in the last ew decades. Sharma (2007) discusses the pipeline integrity regulation requirements (ASME B31.8S, 2014; API RP 580, 2013; API RP 1160, 2013; ASME B31G, 2009, and API 1156, 1999) and how it can be best implemented to achieve reliability, sustainable proitability and regulatory compliance o pipeline systems. Those regulations are not speciically designed or subsea pipeline. Several evaluation codes have been developed or these approaches, such as ASME B31G (2009), Modiied B31G (Szary et al., 2006) PCORRC (Cosham and Hopkins, 2004), DNV-RP-F101 (2010), and Shell 92 (Klever et al., 1995). Because these conventional design codes are based on various assumptions and simpliications, they are not ully able to predict the ailure probability o pipelines, especially when the shape o the structure is more complicated than a simple straight pipe. As a result, the saety actors used in these methods are too high. The calculations o the pipeline lie time and the out o service time are shorter than in reality due to the very conservative nature o the codes. In act, code-based corrosion assessments are mostly probabilistic. Accordingly, the measurements calculated based on the codes are somewhat uncertain and inaccurate and the deterministic methods requently ail to predict the exact burst pressure. The conservative nature o the codes has motivated researchers to select statistical probabilistic methods to obtain more precise and accurate output results. A great deal o attention has ocused on developing probabilistic models that predict the ailure criteria o straight pipelines and their remaining lie time by producing ailure equations. However, very ew studies have examined complicated shape structures such as elbows, U shapes and T shapes. Besides the conventional design codes, numerical analysis methods have been used to evaluate the burst pressure and calculate the remaining strength o elbows with deects. Although the existing research has mainly ocused on the deect size, ew studies have considered the location o the deect. Deects located on extrados exhibit dierent behavior than those located on intrados or the crown area o the elbow. This motivated the authors o this study to develop a new method to achieve more accurate ailure modes or all deect locations on the elbow. Another motivation o this study is to take advantage o the corroded straight pipe ormulas, which were developed using several industrial design codes, to ind an easier and more accurate method or assessing elbows with deects. A quick calculation o the structure lie time and maximum allowable pressure without using Finite Element Analysis (FEA) or computational analyses are the main goals o this study. In addition, the indings o this study are compared with the existing research and methods. Duan and Shen (2006) examined the plastic limit pressure o elbows without deects and with local thinned areas located in the extrados using FEA and experiments. They proposed an empirical ormula or the limit load o elbows with local thinned areas located in the extrados by itting the FEA results and experimentally validated the developed ormula. Li et al. (2001) studied local thin areas and material degradation caused by erosion/corrosion in piping systems and proposed a method to assess the acceptability o the local thin area in an elbow. They then compared the developed method with FEA results. Mohd et al. (2014) examined a straight pipe with a single deect and developed an assessment method by comparing the code-based design data and FEA results. In the present study, the burst pressure o a corrosion damaged elbow was predicted by numerical analysis using ANSYS nonlinear FEA sotware (ANSYS, 2012). The FEA was perormed to prevent uncertainties and inaccuracies in the

3 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~ design codes. The damaged areas in dierent positions o an elbow were modeled using CAD sotware. The burst strength capacity o an elbow with deects was evaluated using empirical models and numerical analyses with both internal and external deects in 10 positions. The FEA results o all o the positions with a single deect were then assessed. A comparison o the numerical analysis and empirical ormulas (industrial codes) was conducted to validate the developed method. The pipeline was then subjected to integrity assessments (i.e., predictions o its structural ailure modes under external and internal pressure). Fig. 1 shows the overall procedure o this study. Service assessment o a corroded elbow Research methods or measuring the corroded pipeline/elbow and select a single corrosion deect size/position Calculate and compare the maximum allowable operating pressure or a s ingle corrosion deected elbow FEM Modiied Goodall ormula Design codes: ASME B31G Modiied ASME B31G DNV-RP-F101 PCORRC Shell 92 Compare the codebased design, Good all ormula, and FEA Conclusion Fig. 1 Overall procedure o the study o a single corrosion damaged steel elbow under internal pressure. The overall procedure or assessing the damaged elbow in this study is summarised as ollows: 1) Research and selection o a single corroded structure (size, material, shape and working environment). 2) Calculation o the burst pressure o a damaged straight pipe using several industrial codes. 3) Burst pressure calculation o the damaged straight pipe using the modiied Goodall ormula. 4) CFD analysis o the internal erosion in the elbow to determine the most deective position. 5) FEA to derive the burst pressure o the erosion/corrosion damaged elbow (modelling, mesh selection, boundary condition, limit load method and evaluation) 6) Comparison o the FEA results with the results o the industrial code-based calculations and the Goodall ormula. 7) Establishment o an improved burst pressure calculation method that is less conservative than the industrial codes and the Goodall ormula (by multiplying the Lorenz actor by the burst pressure o the straight pipe calculated by the industrial codes). 8) Validation o the method and conclusion. DESCRIPTION OF INDUSTRIAL CODE BASED DESIGN ASME B31G In 2009, the American Society o Mechanical Engineers established the ASME B31G model based on a ull-scale burst test o deective straight pipes. The calculation o the remaining strength o pipes with a single deect was suggested by the ASME B31G model together with the prediction o the burst pressure. The ASME B31G design code suggests an evaluation method or the partial metal loss o a pipe wall caused by either internal or external corrosion. The corrosion deect depth

4 438 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~451 varies between 10%~80% o the total wall thickness o the pipe. The longitudinal extent o the corroded area ias covered by the ASME B31G design code. The circumerential extent o the corroded area is disregarded. As shown in Fig. 1, the shape o the corroded area o the pipe wall is idealised to parabolic and rectangular shapes, with the short longitudinal extent idealised as parabolic and the long longitudinal extent idealised as rectangular. The expected ailure pressure is given by Eqs. 1 and 2 or short and long deects, respectively. Short and long deects are deined in the modiied ASME B31G as ollows. I L 50Dt, the deect is assumed to be short. I L 50Dt, then the deect is assumed to be long: P 2d 1 2t SMYS 3t =, (1) D 1 2d 1 M 3t P 2t d = SMYS 1 D t. (2) The bulging stress magniication actor (M) is deined as: M 2 = L. (3) Dt (a) (b) Fig. 2 (a) Parabolic and (b) Rectangular idealisations o a corroded area Modiied ASME B31G The modiied ASME B31G was developed ater the ASME B31G to calculate the strength o the remaining wall thickness o straight pipes ater a deect occurs. The modiied approach uses a calculation method to obtain the Maximum Allowable Operating Pressure (MAOP) in damaged pipes. The ASME B31G pipe design code is generally used or the evaluation o metal wall loss due to either internal or external corrosion. The total deect depth does not exceed 80% o the wall thickness. However, the ASME B31G only deals with the longitudinal extent o the corroded area and does not consider the other circumerential extent. In the modiied ASME B31G, the corrosion damaged area is idealised and assumed to be rectangular in shape. The depth o the idealised rectangular area is taken as 85% o the deepest point o the actual corrosion, as shown in Fig. 3. The short and long deects are deined by the relationship between the length o the deect, the pipe wall thickness and the pipe diameter. Fig. 3 Assumption o the corrosion shape by the modiied ASME.

5 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~ The equation o structural ailure pressure in the modiied ASME B31G diers with the change in the deect length limit. The ailure pressure is estimated by Eqs. (4)-(6). d t P SMYS 69.1MPa t = +. (4) D 1 d M t For L 50Dt (short deect), M is given as: M 2 2 L L = Dt Dt 2. (5) However, i L > 50Dt (long deect), then M 2 = L. (6) Dt DNV-RP-F10 The pipeline with deect model in DNV-RP-F101 is one o the most applicable models or the data used in the oil and gas industries. The model is used to assess single/multiple interacting and complex-shaped deects in pipeline structures. In addition, it introduces an assessment method or single corrosion deects under combined internal pressure and longitudinal compressive stress. The assessment is divided into a saety actor calculation and allowable stress methods. Actual shape o the metal loss deect Assumed rectangular metal loss deect Fig. 4 Assumption o the corrosion shape by DNV-RP-F101. With the saety actor method, a saety actor o 0.9 is used to represent the inaccuracy o the modelled corrosion mass and size. The burst pressure equation is given by Eq. (7) and the allowable stress ailure pressure is given by Eq. (8). The assumption o the corrosion shape according to DNV-RP-F101 is illustrated in Fig. 4. P d 1 2t γ d = γ t m SMTS, (7) D t 1 d 1 γ d Qt d 1 2t P = 0.9F SMTS t. (8) D t 1 d 1 Qt

6 440 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~451 PCORRC The PCORRC model or corrosion damaged straight pipes provides a method or experimentally and numerically calculating the ailure mode. The equation or the ailure mode was developed based on the ailure pressure value o the corrosion deect, assuming that the pipe is composed o ductile material. The ailure mode equation comprises two established ailure predictions: 1) The burst pressure o a plain pipe is the upper limit. 2) The ailure pressure o ininitely long deects is the lower limit. These equations were developed by itting curves to a series o ailure predictions obtained using simple PC sotware. An exponential unction is used to represent the PCORRC burst pressure capacity, which is deined as ollows: P 2t d = 0.95σ u 1 1 e D t L C D ( t d ) 2, (9) σ u where is assumed to be 95% o the ultimate tensile strength o the tensile test and the curve it coeicient (C) is in the case o the conservative prediction o a corroded pipeline. The above equation is then changed to: P 2t d = 0.95s u, Test 1 1 e D t L D ( t d ) 2. (10) Modiied goodall ormula For a thin-walled elbow, Goodall (1978) proposed a ormula or the limit pressure P, which is calculated as: σ t 1 Rm / Rb P0 =. (11) R 1 R / (2 R ) m m b The low stress in the above ormula is deined by σ = ( σ y + σu)/2 as the average stress. The Goodall ormula is used or thin-walled elbows without any damage. Duan and Shen (2006) suggested a modiied ormula or elbows with single deects, shown in Eqs : σ t 1 Rm / Rb PL = ( deect), (12) R 1 R / (2 R ) m m b G G 2 d = , (13) γ θ d G = π /4 2π t. (14) In Section 3, the results o the calculations are compared and validated with the industrial code-based design and numerical analysis results.

7 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~ Elbow burst pressure The assessment o corroded elbows goes beyond the above equations, as with that or piping with pits and local thin areas. This assessment was accomplished in Bubenik and Roseneld s (1993) study Assessing the Strength o Corroded Elbows. In their study, burst tests were conducted on 90 elbows. The results orm the basis o the assessment or elbows with deect(s). The equations or the theoretical elastic stress distributions are presented in the orm o the Lorenz actor, which accounts or the uniorm stress distribution around the circumerence o 90 elbows. Fig. 5 Nominal stress distribution o an elbow (Lorenz actor). The Lorenz actor indicates the increase or decrease in the nominal stress in an elbow relative to a straight pipe, as shown in Fig. 5. The Lorenz actor is also deined in Eq. (15) as ollows: Rb sinα + Rm 2 LF =. (15) Rb + sinα R m For a long radius elbow ( R / R = 3) identical to the target structure, the Lorenz actor reduces to Eqs. (16)-(18). These b m equations calculate the maximum stress distribution o an elbow along the mean angle. Rb 0.5 R m LF = = Rb 1 R m 1.25 ( Intrados), (16) LF = 1 ( Crown), (17) Rb R m LF = = ( Extrados). (18) Rb + 1 R m The range o allowable pressure is determined by applying (1/LF) obtained Lorenz actors to the allowable pressure o a

8 442 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~451 straight pipeline with deect(s), which is obtained rom the calculation o ASME, DNV, PCORRC and the Shell 92 industrial code based design. Table 1 shows the calculation o the burst pressure according to the industrial design codes or a corrosion/ erosion single deect elbow. Table 1 Burst pressure o an elbow with a single deect. BURST PRESSURE ASME B31G (MPa) Modiied ASME B31G (MPa) DNV-RP-F101 (MPa) PCORRC (MPa) Shell 92 (MPa) Modiied Goodall (MPa) INTRADOS CENTERLINE EXTRADOS NUMERICAL SIMULATIONS Target structure An aging subsea gas pipeline elbow was the target structure or this study. The selected elbow had deects in ive internal and ive external positions. The target structure is a pipeline system that located oshore o Terengganu, Malaysia (South China Sea). All o the inormation used in this study, including the size, material and corrosion deect dimensions, were obtained rom Mohd et al. (2014). The geometrical characteristics and material inormation o the target structure (elbow) are outlined in Table 2. Table 2 General inormation o the target structure (elbow). Type o structure Target shape Mean radius (Rm) Outer diameter (D) Bending radius (R b ) Wall thickness (t) Deect dept (d) Deect length (L) γ θ Material grade Gas pipeline 90 elbow (mm) (mm) (mm) 9.5 (mm) 3.9 (mm) 56 (mm) 6 10 API 42 steel In addition to the code-based design o the corrosion/erosion deected elbow, a numerical FEA method was considered to obtain a more accurate prediction o the burst strength capacity o the elbow. With this approach, a three-dimensional elasticplastic numerical analysis using ANSYS was used to simulate the burst/collapse pressure capacity o the deective elbow. Fig. 6 shows the dimensions o the target structure (elbow), including the depth, length and position o the corroded region. Fig. 6 Geometrical coniguration o the corroded elbow.

9 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~ Fig. 7 Corroded region depicted by CFD analysis. Predicting erosion by Computational Fluid Dynamics (CFD) is a three-step process. The irst step is to model the low-ield using the Eulerian approach. In the second step, particle tracking is perormed using the Lagrangian approach. In this step, each particle is considered as a discrete entity and the particle trajectory is calculated based on the calculated low-ield in the irst step and the exchange orces. Finally, the data provided in the second step are used to calculate the erosion rate at the wall. The conservation equations or mass and momentum are written as ollows:.( ρu) = 0, (19).( ρ uu) = p +.( µ u) S. (20) E x Eq. (21) is the momentum equation in the X direction. The momentum equations or the Y and Z directions are similar. The term S E in the momentum equation represents the momentum exchange between the continuous low-ield and the particles. Fluent 6.3 was used to solve the governing equations with the inite volume method. The standard k-epsilon model and the standard wall unction were used to model the turbulence eects. The trajectory o each particle was calculated by integrating the orce balance on the particles. The governing equation o the motion o the particles is written as ollows: du p mp = Fd( u up) + Fb + Fp + FM, (21) dt where p and denote the particle and low, respectively. The irst term on the right hand side represents the drag orce and the other terms represent the buoyancy orce, pressure gradient orce and virtual mass orce, respectively. Two-way coupling was used to model the interactions between the particles and the continuous low ield. Ater ollowing the above procedures, the impingement data, such as the speed and the angle o impact, are provided as the particles hit the wall. Using this inormation, the erosion rate can be calculated. The erosion rate is deined as ollows: R Erosion N Particles ( ) mcd ( ) ( ) bv p p a up =, (22) A p= 1 aces where m p is the mass low rate o the particles, u p is the velocity o particles and A aces is the area o the grid cell. Previous studies have revealed that the relected velocity o the particles is less than the incoming velocity. Furthermore, the angle o impact has been observed to have a signiicant eect on the coeicients o restitution. The perpendicular and parallel coeicients o restitution or sand impacting API X42 carbon steel are incorporated into the model. e θ θ θ θ per = , (23) e θ θ θ θ θ par = , (24)

10 444 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~451 where θ is the impact angle o the particles, the impact angle unction (α) is a piecewise linear unction and the diameter unction C(d p ) and the velocity exponent unction b(v) are the model constants. The carrier luid in this study was natural gas and the mass low rate o the sand particles was speciied as equal to 0.05 g/s at the inlet. The velocity o natural gas at the inlet was 10 m/s and the diameter o the sand particles was 0.04 inches. An elbow with an internal diameter o mm was considered. A grid consisting o approximately 326,500 hexahedral cells was generated using Gambit Fig. 7 shows the grid and the predicted trajectory o the sand particles. The CFD simulation method used in this study was identical to the method used in the application and experimental validation o the CFD-based erosion prediction model or elbows and plugged tees in Chen et al. (2004). Fig. 8 shows the comparison o the predicted data by simulation with the experimental results. Fig. 8 Validation o the CFD analysis prediction with experimental results. Material API X42 (2007) is commonly used to manuacture the pipelines used in the oil and gas industries. The target structure installed oshore o Terengganu, Malaysia, was also made rom API X42. The material properties o API X42 used in the FEA are shown in Table 3. Fig. 9 shows the true stress-strain curve o the API X42 steel. In this study, the true stress-strain experimental data based on Cronin (2000) were used or the FEA. Table 3 API X42 steel material properties. Young s Modulus (GPa) E Poisson s Ratio ν Yield strength (MPa) σ y Tensile strength (MPa) σ u Fig. 9 True stress-strain curve o API X42 steel.

11 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~ FE Modelling The results o the CFD analysis showed that erosion corrosion appeared mostly at the centre o the bending angle, as shown in Fig. 7. Five (two positions on the extrados, two positions on the intrados and one position on the centre line) external deect positions and ive (two positions on the extrados, two positions on intrados and one position on the centre line) internal deect positions were modelled to more accurately observe the behaviour o the elbow, as shown in Fig. 6. The corrosion/erosion at these positions was idealised to a rectangular shape to provide the appropriate conditions or comparison with the code-based design results. Full scale eight node iso-parametric brick (Solid 185) elements with a reduced integration option model were used or the target structure (elbow). The structure was irst modelled by 3D CAD modeller and then exported to ANSYS or numerical simulation. An inelastic multi-linear material model was also used or the structure. Mesh and element size A set o mesh convergence tests or several cases were perormed to determine the appropriate size and quantity o mesh. It is desirable to ind the minimum number o elements that give a converged solution. Fig. 10 shows the mesh convergence test results. Fig. 10 Mesh convergence results. According to the mesh convergence tests, 41,536 elements and 51,288 nodes were selected with the symmetry condition, including the corrosion deect areas or all deect positions, shown in Fig. 11. Fig. 11 Corrosion positions and idealised conigurations o the deect structure.

12 446 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~451 Several internal and external deects located at dierent degrees o mean radius were perormed with the FE model, as shown in Fig. 12. The size and shape o the damaged area were determined according to the industrial code assumptions. (a) (b) Fig. 12 FE model o the internally (a) and externally (b) corroded elbow. Loads and boundary conditions Internal pressure was added to all internal aces o the elbow to perorm the burst strength test. For the boundary condition, both aces were bounded as UX=UY=UZ to restrict the movement o the elbow to the required directions during the FEA. Fig. 13 shows the boundary condition and internal load application or the FEA. Fig. 13 Load and boundary condition o the corroded/eroded elbow. Load limit deinition An ideal limit load, when the load corresponds to the limit state, occurs when the load stops increasing but the strain rate o the displacement continues increasing to ininity. The hypothesis o the load limit deinition is that the material o the structure is assumed to be elastic and perectly plastic material with only small displacements. Fig. 14 Determination o the collapse load using the Twice- Elastic-Slope (TES) line or plastic instability point.

13 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~ In reality, such an ideal material does not exist because o strain hardening and geometry hardening or weakening. Accordingly, several load limit value determining methods have been proposed in the engineering ield, such as the Twice- Elastic-Slope (TES) method, three-times-elastic-slope method, twice-elastic-deect method, tangent-intersection method, zerocurvature method and 0.2% residual-strain method. In this study, the limit load determination is based on the load-strain curves using the TES method. The TES method is described in detail in the ASME Boilers and Pressure Vessel Code (2010). The strain in the load-strain curve is the maximum von-mises strain o the elbow. RESULTS AND DISCUSSION The numerical analysis o the burst pressure or all positions (1~10) was conducted using FEA. The burst pressure o the elbow with deects was also calculated with the modiied Goodall ormula introduced in the previous section. A new method was proposed to calculate the burst pressure o an elbow with a single deect using the industrial codes and the Lorenz actor. The burst pressure o a straight pipe with a single deect was irst calculated and the inverse Lorenz actor (1/LF) was then multiplied by the results or the straight pipe burst pressure to achieve the burst pressure o the elbow with deects. Table 4 compares the burst pressures o the elbow with a single deect calculated by FEA, the Goodall ormula and the industrial codebased method. Table 4 Burst pressure calculation results based on industrial codes, the Goodall ormula and FEA. The comparison o the FEA results with those o the industrial code and Goodall ormula showed that errors existed. The maximum error in the comparison o the FEA results and the Goodall ormula was 33.0 percent and the minimum error was 18.3 percent. The maximum and minimum errors in the comparison o the FEA and ASME B31G results were 29.9 and 14.1 percent, respectively. This code was then upgraded to the modiied ASME B31G, which is less conservative, and the respective errors reduced to 23.9 and 7.3 percent, respectively. Shell 92 appeared to be even less conservative than the modiied ASME B31G. The errors in Shell 92 reduced to a maximum and minimum o 20.3 and 3.1 percent, respectively. With the DNV-RP- F101 and PCORRC codes, the results were more comparable with the FEA results. The maximum and minimum errors were 15.4 and 0.3 percent or PCORRC and 12.5 and 1.3 percent or DNV-RP-F101, respectively. The results shown in Table 4 are illustrated in Figs. 15 and 16, which compare the calculation results or the burst pressure o the elbow with deects based on industrial codes and the Goodall ormula.

14 448 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~451 Fig. 15 Comparison o the burst pressure calculation results base on industrial codes and the Goodall ormula. Fig. 16 Comparison o the elbow burst pressure calculated by ASME B31G, modiied ASME B31G, DNV-RP-F101, PCORRC, Shell 92, FEM and the modiied Goodall ormula. Fig. 16 illustrates the burst pressure o the elbow with deects calculated by ASME B31G, modiied ASME B31G, DNV- RP-F101, PCORRC, Shell 92, FEM, and the modiied Goodall ormula. Fig. 16 also shows the curve trend and the gap between the FEA results and the proposed industrial code-based calculation method and the Goodall ormula. As shown in Figs. 16 and 17, the modiied Goodall ormula was the most conservative method. The curve appears to be similar to a straight line, so the dierent stress distributions o the extrados and intrados o the elbow with deects were not taken into consideration. The hatched area between the two lines (the FEA and Goodall ormula results) shows that the stress distribution o the elbow was not considered (Fig. 17(a)). The calculation o the burst pressure with the ASME B31G code illustrated in Fig. 17(b) shows that the curvature o the FEA results and ASME B31G code ollow a similar trend. However, the gap between the two lines is big, which means the ASME B31G code was highly conservative. This code was then upgraded to the less conservative modiied ASME B31G code. Fig. 17(c) illustrates the curvature and gap between the FEA and modiied ASME B31G results. The Shell 92 code-based calculation results are compared with the FEA results in Fig 17(d). The curvature o the two lines appears similar to that o the modiied ASME B31G, although the gap between the lines is reduced. The results o the calculations based on the PCORRC and DNV-RP-F101 codes are illustrated in Fig. 17(e) and (). The curvature o the FEA and code-based calculation results are very similar and the gaps between the two lines are very small.

15 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~ Fig. 17 Curvatures and gaps between the FEA results and ASME B31G, modiied ASME B31G, DNV-RP-F101, PCORRC, Shell 92, and modiied Goodall ormula based calculation results. As shown in Fig. 17, the trend o the burst pressures o the damaged elbows calculated by the industrial codes was identical to that o the FEA results, in contrast to the modiied Goodall ormula. The FEA results show that the burst pressure o the elbow varied not only by deect size but also the position o the deect around the mean radius. The result curves show the reduction in the burst pressure o the elbow with the same size o deect but dierent positions. The extrados had the lowest stress distribution and the intrados had the highest stress distribution. Elbows with deects on the intrados thereore ail earlier than those with deects on the extrados. Multiplication o 1/LF by the burst pressure o the straight pipe calculated by the industrial codes gave a burst pressure trend that was similar to the FEA and less conservative than the modiied Goodall ormula. Overall, the developed method provides a simple calculation o the burst pressure o elbows with deects that can be used instead o FEA. burst pressureo straigth pipe with deect calculated by industrial code 1 = LF burst pressureo elbow with deect. (25) Eq. (25) is a simple equation or calculating the burst pressure o elbows with deects using industrial codes. CONCLUSION In this study, a local ailure criterion or API X42 steel was used to predict the ductile ailure o ull-scale pipe elbows with simulated corrosion/erosion under internal pressure. The local ailure criterion was the stress-modiied racture strain

16 450 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~451 or normalised API X42 steel as a unction o the stress triaxiality (deined by the ratio o the hydrostatic stress to the eective stress). For a pipe elbow with simulated corrosion/erosion deects, the results o the FEA with the proposed local racture criterion indicated that predicted ailure took place ater the deective pipe elbow attained maximum loads or all cases, and thus the present approach suggests that pipe elbow ailure is governed by global instability. A parametric study was perormed, rom which a simple method (multiplication o 1/LF by the burst pressure o the elbow according to the deect position on the mean angle) is proposed to predict the burst pressure or structural deects in gas pipe elbows made o the particular API X42 steel considered in the present work. The burst pressure capacity o a damaged straight pipe was also calculated according to several industry codes (ASME B31G, modiied ASME B31G, DNV-RP-F101, PCORRC and Shell 92). The burst pressures o the damaged straight pipe were then used to calculate the burst pressure capacity o an elbow with an identical area o damage. The results were compared with those o the modiied Goodall ormula. Based on the results o the code-based design, modiied Goodall ormula and FEA, the ollowing actors can be concluded: The burst pressure predictions using the codes varied considerably in the straight pipeline and curved (elbow) structures. The burst pressure o the elbow with deects located on the intrados appeared lower than that o the straight pipe. However, the burst pressure o the elbow with deects located on the extrados appeared higher than that o a straight pipe with an identical deect size. The burst pressure o a damaged elbow could be simply determined by multiplying the Lorenz actor by the calculated burst pressure o the damaged straight pipe rom the industrial codes, which was more accurate than the modiied Goodall ormula. The modiied Goodall ormula was the most conservative method and the burst pressures calculated by the modiied Goodall ormula were not consistent with the FEA results or the code-based design trend (extrados, crown, intrados) All o the industrial code based calculations o the damaged elbow were conservative compared with the FEA results, except or DNV-RP-F101. However, the results calculated by the multiplication o 1/LF by the industrial code results were more comparable with the FEA results than those calculated by the modiied Goodall ormula. The results calculated by the modiied Goodall ormula appeared very conservative. The modiied Goodall ormula was developed based on the burst pressure o weakest point (intrados) o the elbow and other deect locations such as extrados or crown were not considered. The FEA method provides a general and reliable way to assess the burst pressure o complex corrosion deect shapes. Some codes such as DNV-RP-F101 oer guidelines or using FEA at an acceptable saety level. ACKNOWLEDGEMENTS This research was conducted under the project to establish the oundations o industrial technology which is unded by the Ministry o Trade, Industry & Energy (MOTIE, Korea) (Grant no.: G02N ) REFERENCES ANSYS, ANSYS user manual (Release 14.0). Canonsburg, PA: ANSYS Inc. API, Speciication or line pipe. Washington: American Petroleum Institute. API 1156, Eects o smooth and rock dents on liquid petroleum pipelines, Phase I and Phase 2. USA: American Petroleum Institute. API RP 1160, Managing system integrity or hazardous liquid pipelines. USA: American Petroleum Institute. API RP 580, Recommended practice or Risk-Based inspection. USA: American Petroleum Institute. ASME BPVC, Rules or construction o pressure vessels. Boiler and Pressure Vessel Code Section VIII division 2. New York: American Petroleum Institute. ASME B31G, Manual or determining the remaining strength o corroded pipelines. a supplement to ASME B31G code or pressure piping. New York: American Petroleum Institute. ASME B31.8S, Managing system integrity o gas. ASME code or pressure piping, B31 supplement to ASME B31.8. New York: ASME.

17 Int. J. Nav. Archit. Ocean Eng. (2015) 7:435~ Bai, Y. and Bai, Q., Subsea pipelines and risers. 1st Ed. MA: Elsevier Science. Bai, Y. and Bai, Q., Subsea pipelines integrity and risk management. 1st Ed. MA: Elsevier Science. Bubenik, T.A. and Roseneld, M.J., Topical report on assessing the strength o corroded elbows. Columbus: Battelle press. Chen, E., Mclaury, B.S.M and Shirazi, S.A., Application and experimental validation o a computational luid dynamics (CFD)-based erosion prediction model in elbows and plugged tees. Computers & Fluid, 33(10), pp Cosham, A. and Hopkins, P An overview o the pipeline deect assessment manual (PDAM). 4th International Pipeline Technology Conerence, Ostende, Belgium, 9-13 May Cronin, D.S., Assessment o corrosion deects in pipelines. Ph.D. thesis. University o Waterloo, Waterloo, Canada. DNV, Corroded pipelines ( DNV-RP-F101), Oslo: Det Norske Veritas. Duan, Z.X. and Shen, S.M., Analysis and experiments on the plastic limit pressure o elbows. Nanjing: College o Mechanical and Power Engineering, Nanjing University o Technology. Goodall, I.W., Lower bound limit analysis o curved tubes loaded by combined internal pressure and in-plane bending moment. (CEGB RD/B/N4360). UK: Central Electricity Generating Board. Kim, D.W., Mohd, H.M., Lee, B.J., Kim, D.K., Seo, J.K., Kim, B.J. and Paik, J.K., Investigation on the burst strength capacity o aging subsea gas pipeline. ASME nd International Conerence on Ocean, Oshore and Arctic Engineering, Nantes, France, 9-14 June Klever, F.J., Stewart, G. and Valk, C.A.C., New developments in burst strength predictions or locally corroded pipelines. The Proceedings o the 14th International Conerence on Oshore Mechanics and Arctic Engineering (OMAE), Copenhagen Denmark, June Kyriakides, S. and Corona, E., Mechanics o oshore pipelines-volume1: buckling and collapse. MA: Elsevier Science. Li, Z., Yinpei, W., Jin, C. and Cengdian, L., Evaluation o local thinned pressurized elbow. International Journal o Pressure Vessels and Piping, 78(10), pp Mohd, M.H., Kim, D.W., Lee, B.J., Kim, D.K., Seo, J.K. and Paik, J.K On the burst strength capacity o an aging subsea gas pipeline. Journal o Oshore Mechanics and Arctic Engineering, 136(4), pp.1-7. Sharma, P.P, Pipeline integrity management or sustainable proitability and compliance. Rio pipeline conerence & exposition 2007, Rio de Janeiro, 2-4 October Szary, T., The Finite element method analysis or assessing the remaining strength o corroded oil ield casing and tubing. PhD thesis. Mechanical Engineering, Germany.