Engineering Structures 33 (2011) 25472553

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Propagation buckling in deep sub-sea pipelines

F. Albermani , H. Khalilpasha, H. Karampour
School of Civil Engineering, The University of Queensland, 4072, Australia

article

info

abstract
The paper investigates buckle propagation in deep sub-sea pipelines. Experimental results are presented using ring squash tests and hyperbaric chamber tests, and are compared with a modified analytical solution and with numerical results using finite element analysis. The experimental investigation was conducted using commercial aluminium pipes with diameter-to-thickness (D/t ) ratio in the range 2048. In contrast to conventional cylindrical pipe, a faceted cylindrical geometry is also investigated. Preliminary analysis of a faceted pipe shows that a substantial increase in buckling capacity can be achieved for the same D/t ratio. Crown Copyright 2011 Published by Elsevier Ltd. All rights reserved.

Article history: Received 8 October 2010 Received in revised form 20 April 2011 Accepted 27 April 2011 Available online 1 June 2011 Keywords: Buckle propagation Sub-sea pipelines

1. Introduction The recent failure of Deepwater Horizon in the Gulf of Mexico (April 2010) and the resulting oil spill (40,00060,000 barrels/day) that took months to contain is a reminder of the risk involved in deep sub-sea hydrocarbon exploration. Another recent spill that also took months to contain was the Montara West Atlas rig in Australia, in August 2009. The environmental and economic impact of these catastrophes is substantial and will take many years to quantify. Most of the available hydrocarbon reserves are located in remote ultra-deep sub-sea regions (over 1500 m depth) and exploration in such regions poses many engineering challenges. Hence, it is vital to develop engineering solutions that will allow safe and economical realization of these resources. A sub-sea pipeline can experience a number of structural instabilities, such as lateral (snaking) buckling, upheaval buckling, span formation and propagation buckling. Among these, propagation buckling is the most critical one, particularly in deep water, and can quickly damage many kilometres of pipeline. A local buckle, ovalization, dent or corrosion in the pipe wall can quickly transform the pipe cross-section into a dumb-bell (or dog-bone) shape that travels along the pipeline as long as the external pressure is high enough to sustain propagation. The lowest pressure that maintains propagation is the propagation pressure which is only a small fraction of the elastic collapse pressure of the intact pipe. This results in a substantial increase in the material and the installation cost of the pipeline, since design is therefore governed by propagation pressure.

Many researchers have investigated various aspects of this problem since it was first presented by Mesloh et al. [1,2] and Palmer and Martin [3]. Most notably is the extensive work of Kyriakides (for example [46]) and Calladine [7,8]. Recent books by Kyriakides [9] and Palmer and King [10] provide comprehensive review of this problem and the associated literature. Xue and Hoo Fatt [11] investigated buckle propagation in corroded pipelines. A number of solutions to guard against buckle propagation (and the resulting increase in the wall thickness of the pipeline) were proposed such as buckle arrestors [9], pipe-in-pipe system [6], sandwich pipe system [12] and ring-stiffened pipelines [13]. This paper will first propose a modification to the lower bound solution presented by Palmer and Martin [3]. A summary of experimental results using ring squash tests as was proposed by Kamalarasa and Calladine [7] will follow. Experimental results conducted on 3 m long pipes in a hyperbaric chamber and finite element analysis results verified against the experimental results will also be presented. Finally, a new pipeline design will be proposed that has the potential to increase the propagation buckling capacity without increasing the wall thickness of the pipeline. It is expected that this new design will have additional benefits in alleviating other possible instabilities that may take place in the pipeline. 2. Analytical solution of propagation pressure A number of empirical formulae have been proposed to estimate propagation pressure as a function of D/t (diameter to wall-thickness ratio) and the material yield stress (e.g. [1,2]). However, the first analytical solution for propagation pressure in a sub-sea pipeline was presented by Palmer and Martin [3]. When compared with experimental results, the Palmer and Martin (PM) solution underestimates the propagation pressure.

Corresponding author. E-mail address: f.albermani@uq.edu.au (F. Albermani).

0141-0296/$ see front matter Crown Copyright 2011 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.engstruct.2011.04.026

2548

F. Albermani et al. / Engineering Structures 33 (2011) 25472553

c
Fig. 1. A schematic of deformation stages in propagation buckling.

This difference increases as D/t decreases. The PM solution is based on a ring collapse (plane strain) and accounts for the circumferential bending effect of the pipe wall (Fig. 1(a)(b)). Many researchers have attempted to modify the PM solution by including a strain hardening effect [14], a surface stretching effect [8] and by using an energy balance approach [15]. The propagation pressure from the PM solution, PPM , is given by (see also Fig. 1(a)(b)): ppm =

2
t r (1)
Fig. 2. Ring squash test RST.

for a pipe with radius, r , wall thickness, t , and material yield stress, y . Based on experimental observations from hyperbaric chamber and ring squash tests, the top and bottom hinges (shown as T and B in Fig. 1(a)) move towards each other while the left and right hinges (L and R in Fig. 1(a)) move laterally away from each other. This deformation continues until touchdown (when contact is established between T and B, Fig. 1(b)), the lateral movement seizes and flattening of the resulting four arch segments commence (Fig. 1(c)). Accordingly, we propose a modification to the lower bound PM solution, by accounting for the circumferential membrane as well as flexural effects in the pipe wall (Fig. 1(a)(c)): Wex = (Win )f + (Win )m (2)

The propagation pressure predicted by Eq. (7) is 19% higher than the PM prediction, regardless of t /r ratio. Furthermore, by invoking plane strain conditions, the tensile coupon yield stress can be amplified by a factor of (2/ 3) in Eq. (7) that results in an . additional 15% increase in p 3. Ring squash tests (RST) Kamalarasa and Calladine [7] proposed a simple ring squash test (RST) for approximating the propagation pressure. The RST is performed by diagonally compressing a short segment, L, of a pipe (L = 150 mm is used in this study) between two rigid indenters of the same radius as the pipe being tested. Fig. 2 shows a typical RST conducted in this study. The RST can also be used to calculate an effective yield stress which implicitly accounts for the strain hardening response of the material under propagation pressure. A quarter section of the pipe under RST is depicted in Fig. 3. Using equilibrium, the effective yield stress can be calculated from

where Wex is the external work done by the net hydrostatic pressure and Win is the internal work due to circumferential flexure, f , and membrane, m, effects. The initially circular crosssection of the pipe (Fig. 1(a)) will deform into a dog-bone (Fig. 1(b)) and eventually into a nearly flat segment (Fig. 1(c)). Accordingly, Eq. (2) can be written as p(A) = 3 mp + (pr )(l) (3) where A is the change in the cross-section area (from a to c in Fig. 1), l is the change in the circumferential length and mp is the plastic moment, these are given by

y = (Fo r )/(Lt 2 )

(8)

A = r 2 l = 0.626r
mp = y t
2

(4) (5)

where Fo is the load level at which plastic hinges form in the ring test. Using an energy balance approach, an estimate of the propagation pressure based on RST, PRST , can be obtained from calculating the area, U , under the experimental loaddisplacement curve and the change in the specimen cross-section area, A, as shown in Fig. 4: PRST (A) = U /L. (9)

. (6) 4 , is Substituting Eqs. (4)(6) into (3), the propagation pressure, p obtained as = p
3 2.515

2
t r

= 1.193ppm .

(7)

Aluminium pipes (commercial pipes, alloy 6060) were used in this study (Table 1). Three different nominal outer diameters, D, were used ranging from 50 to 76 mm with nominal wall thickness, t , ranging from 1.6 to 3 mm (D/t = 2047.5). Two RST (A and B) were conducted for each pipe diameter (a total of 6 RST). A summary

F. Albermani et al. / Engineering Structures 33 (2011) 25472553 Table 1 Summary of experimental, analytical and numerical results. Sample/material ID D50 D60 D76 Al-6060 T591 T4 T5 25 20 47.5 D/t Coupon tests Ring tests (MPa) E / y 440 716 367 E /E (%) 1.5 1.9 0.4 Analytical (MPa) PPM Hyperbaric chamber (MPa) Experiment PI 6.42 8.24 1.32 PP 1.6 2.3 0.35 Finite element PIFE 5.12 8.15 1.07 PPFE 1.1 1.6 0.3

2549

p
Eq. (7) 0.93 1.21 0.245

y (MPa)
122 81 156

y Eq. (8)
170 143 164

PRST Eq. (9) 1.045 1.67 0.314

Eq. (1) 0.778 1.011 0.205

Table 2 Comparison of experimental, analytical and numerical results. Sample D/t Ring test Hyperbaric chamber Finite element PP /PRST 1.531 1.377 1.115 PI /PP 4.01 3.58 3.77 PI /PIFE 1.253 1.011 1.234 PP /PPFE 1.453 1.437 1.167

PRST /p
D50 D60 D76 25 20 47.5 1.124 1.380 1.282

PP /p
1.720 1.900 1.428

Fig. 3. A quarter model of the RST.

of the average yield stress and propagation pressure (Eqs. (8) and (9)) obtained from the RST together with results from longitudinal tensile coupon tests are given in Table 1. Fig. 5 compares the deformed pipes at the end of the RST to the initial intact section. It is worth noting that the RST is terminated at touchdown (Fig. 2(b)), however when the load is released elastic spring-back takes place. This explains the gap in the dumb-bell shapes shown in Fig. 5. Fig. 6 shows a typical experimental loaddisplacement curve from RSTs for D60 A and B samples. From the RST experimental loaddisplacement curve, the area under this curve, U , is calculated. This area is composed of elastic and plastic components. Similarly, from the deformed RST specimens (Fig. 5), the change in the cross-section area A is calculated. Using U and A, an estimate of PRST is obtained from Eq. (9). In this work, PRST is calculated twice; using total U and using only the plastic component of U to account for the elastic springback. The experimental results for PRST are normalized by PPM (Eq. (1)) and presented in Fig. 7 against nominal D/t . Each point in this figure is the average of two test samples. Also shown in the figure is the modified solution from Eq. (7). The average ratio of for all the tests is 1.262 (Table 2). PRST /p In order to investigate the viability of substituting the physical test (RST) by numerical simulation, a thin shell finite element model (one quarter model) using material properties obtained from tensile coupon specimens was used. From the FE analysis, the deformed shape of the pipe cross-section and the loaddeflection curve (Fig. 8) are obtained for each test and used to calculate U and A, and hence PRST (Eq. (9)). The FE prediction of PRST (normalized by PPM ) is given in Fig. 7. Although the FE simulation provides

Fig. 4. Typical loaddisplacement response in RST.

reasonable results, the RST remains a simple and expedient test to conduct. 4. Propagation pressure using quasi-static tests in the hyperbaric chamber A stiff 4 m long hyperbaric chamber rated for 20 MPa (2000 m water depth) internal pressure was used for testing (Fig. 9). Three metre long aluminium pipes similar to the ones used in the RSTs (L/D = 4060) were used in the hyperbaric chamber tests. Ovalization measurements along the pipe samples before testing were carried out that gave an average ovalization ratio (Eq. (10)) around 0.46%0.67%

Dmax Dmin Dmax + Dmin

(10)

where Dmax and Dmin are the maximum and minimum measured outer diameters along the pipe length.

2550

F. Albermani et al. / Engineering Structures 33 (2011) 25472553

Fig. 5. Comparison of intact and deformed samples from RST.

Fig. 6. Experimental loaddisplacement curve from RST for D60.

Fig. 8. Comparison of FE and experimental loaddisplacement response of RST for D76.

Fig. 7. Comparison of experimental, analytical and numerical results from RST.

The intact 3 m long pipe was sealed at both ends by welding on a thick disc, filled with water and placed inside the hyperbaric chamber. The pipe specimen inside the chamber was vented to the atmosphere by a small tube projecting through the chamber wall. The chamber was sealed, filled with water and pressurized at a slow rate using a high pressure pump. When the pressure reached a certain value, the initiation pressure PI , a section along the pipe sample collapsed leading to a significant drop in chamber pressure and water flowed from within the pipe sample through the vent. By maintaining low rate of pumping, the chamber pressure is stabilized at the propagation pressure, Pp , with the buckle longitudinally propagating along the pipe sample accompanied by uniform water flow from the vent. Fig. 10 shows some of the specimens after testing in the hyperbaric chamber. Table 1 gives a summary of average results obtained from the hyperbaric chamber tests (total of 19 tests). A typical pressurevolume change response from the hyperbaric chamber

Fig. 9. The hyperbaric chamber used in this study.

tests is shown in Fig. 11. In this figure, the pressure is normalized by PPM and the change in the pipe volume V is normalized by the initial volume of the intact pipe, V . As a snap-through buckling problem, PI is expected to be sensitive to imperfections (such as a dent in the pipe wall) though the pressure will be stabilized at Pp regardless of imperfections. FE simulation of 3 m long pipe samples used in the hyperbaric chamber tests were conducted using ANSYS 12.1 [16]. Thin 4-noded shell elements (181) were used to model the pipe and frictionless contact and target elements (174 and 170) were used to define the contact between the inner surfaces of the pipe wall. Due to symmetry, a one half model of the pipe wall

F. Albermani et al. / Engineering Structures 33 (2011) 25472553

2551

FE model. A von Mises elastoplastic material definition with isotropic hardening was adopted. The stressstrain curve obtained from tensile longitudinal coupons taken from the pipe wall was used. The initiation and propagation pressures obtained from FE analysis (PIFE and PPFE respectively) are summarized in Table 1 and are in reasonable agreement with the experimental results from the hyperbaric chamber. A typical FE result for 50 2 mm pipe is shown in Figs. 11 and 12. Fig. 13 shows the FE predicted deformations of the same pipe at four stages of buckle propagation (V /V = 0.02, 0.05, 0.07 and 0.1). Based on V /V = 0.1 in Fig. 13, the length of the transition zone, , is 300 mm (the difference between z = 1500 and 1200 mm) which gives /D = 6. The length of the transition zone was stipulated to be around 10D by Chater and Hutchinson [15], while Kamalarasa and Calladine [8], using dimensional analysis, indicated that the length is given by

Fig. 10. 3 m long pipe samples after testing in the hyperbaric chamber.

= 3.6

r3 t

(11)

which gives 318 mm for 50 2 mm pipe. 5. Discussion Table 2 gives a comparison of the analytical, numerical and experimental results presented in the previous sections. The ratio of propagation pressure from the hyperbaric chamber tests PP (Eq. (7)) vary from 1.428 to the modified analytical solution p to 1.9 depending on D/t ratio. This is in agreement with the empirical relation proposed by Mesloh et al. [2]. Similarly, the ratio of propagation pressure from the hyperbaric chamber tests PP to the ring squash tests PRST (Eq. (9)) vary from 1.115 to 1.531 with an average of 1.341, close to 1.4 suggested in [7]. The FE predictions of initiation and propagation pressures on average represent 87% and 74%, respectively, of the experimental results. The FE results are based on material properties from tensile longitudinal coupons (Table 1) taken from the pipe wall (since such material properties are readily available to the analyst). A better result of PPFE (particularly for D/t 20 and 25) would be obtained if yield stress based on the ring test (Table 1) rather than coupon test is used in the FE model. The ratio of PI /PP from the hyperbaric chamber tests is also shown in Table 2 and varies from 3.5 to 4.0. These results highlight the vulnerability of deep sub-sea pipelines to propagation buckling that can quickly spread and damage long sections of the pipeline. One of the solutions used to guard against this form of buckling is the use of buckle arrestors at various locations along the pipeline. The effect of buckle arrestors is to locally stiffen the pipe circumference and hence contain possible collapse within the pipe to segments between two arrestors. Miura [17,18] has shown that the post-buckling configuration of an axially loaded cylindrical shell offers some useful structural characteristics and in particular high circumferential flexural rigidity. In order to exploit this rigidity, a faceted cylindrical pipe as shown in Fig. 14(a) is proposed instead of a conventional cylindrical pipe of the same D/t ratio. Other faceted configurations are also possible and should be explored. Issues such as production of such new faceted configurations and any associated flow losses will not be discussed here. A six sided faceted pipe, 3 m long with an overall diameter of 50 mm and wall thickness of 2 mm is proposed to be used for comparison with the experimental and FE results presented in the previous section for a cylindrical aluminium pipe (Fig. 11). An isosceles right triangle is proposed to be used with a modular facet which gives an amplitude of undulation of 0.134r . It is envisaged that the faceted geometry will converge to smooth cylindrical geometry at both ends of the pipe to facilitate welding to adjacent

Fig. 11. Normalized pressurevolume response (experimental and numerical results) for D50.

Fig. 12. Deformation contours for the central part of 3 m long D50 test showing propagation.

(180) was discretized using 750 elements along the length and 20 elements in the circumferential direction, with 5 integration points through the thickness. In order to control the nonlinear analysis, a 1% ovalization was introduced at mid-length in the

2552

F. Albermani et al. / Engineering Structures 33 (2011) 25472553

L = 3000 mm

Fig. 13. Predicted deformed shapes at various stages of propagations.

pipe segments. It is believed that this can be achieved as the two geometries are isometric. An FE model of the faceted pipe has been generated using ANSYS 12.1 [16]. The same material properties for D50 cylindrical pipe (Table 1) were used. Quad shell 181 elements were used and a full model was generated for the pipe as shown in Fig. 14(b). Due to the inherent faceted configuration, there is no need to introduce any small initial imperfection to control the nonlinear analysis as was done in the previous section. The faceted pipe response under external pressure is shown in Fig. 11. It is clear from this figure that the faceted geometry should result in a 76% increase in initiation pressure and 127% increase in propagation pressure for the same D/t ratio. Following the DnV [19] guidelines, for example, this increase in buckling capacity for the same D/t ratio can be translated to around 22% material saving. 6. Conclusion Experimental results for buckle propagation in pipelines are presented. Results using quasi-static tests in a hyperbaric chamber on 3 m long aluminium pipes as well as results from ring squash tests are summarized. Based on the experimental observations, a modified analytical solution for buckle propagation is proposed. Nonlinear finite element analysis is conducted and verified against the hyperbaric chamber tests. In general, reasonable agreement is obtained between analytical, numerical and experimental results. In order to increase propagation buckling capacity of a pipeline without increasing the wall thickness, faceted pipe geometry is proposed. Preliminary result from finite element analysis of the proposed faceted pipe indicates a substantial increase in initiation and propagation buckling capacities for the same D/t ratio

Fig. 14. (a) Proposed faceted pipe configuration, (b) FE model of the faceted pipe.

F. Albermani et al. / Engineering Structures 33 (2011) 25472553

2553

(in comparison to conventional cylindrical pipe). This could result in great saving in material and construction cost. It is expected that the faceted geometry will have additional benefits in alleviating imperfection sensitivity and other possible instabilities that may take place in the pipeline. Further investigation is in progress to fully explore faceted pipe geometry. Acknowledgements This research is being undertaken within the CSIRO Wealth from Oceans Flagship Cluster on Subsea Pipelines with funding from the CSIRO Flagship Collaboration Fund. We would like to acknowledge the technical assistance provided by the technical staff at the Structures laboratory in the School of Civil Engineering at the University of Queensland. References
[1] Mesloh RE, Sorenson JE, Atterbury TJ. Buckling and offshore pipelines. Gas Mag 1973;7:404. [2] Mesloh RE, Johns TG, Sorenson JE. The propagating buckle. In: Proceedings international conference behaviour of offshore structures. BOSS. vol. 1. 1976. p. 78797. [3] Palmer AC, Martin JH. Buckle propagation in submarine pipelines. Nature 1975;254(5495):468. [4] Kyriakides S, Babcock CD. Experimental determination of the propagation pressure of circular pipes. ASME, J Press Vessel Technol 1981;103:32836.

[5] Kyriakides S, Netto TA. On the dynamics of propagating buckles in pipelines. Internat J Solids Structures 2000;37(4647):684368. [6] Kyriakides S, Vogler TJ. Buckle propagation in pipe-in-pipe systems. Internat J Solids Structures 2001;39(2):36793. [7] Kamalarasa S, Calladine CR. Buckle propagation in submarine pipelines. Int J Mech Sci 1988;30(34):21729. [8] Kamalarasa S, Calladine CR. Geometry and strain in the transition region of a collapsing submarine pipe. Int J Mech Sci 1989;31(3):20719. [9] Kyriakides S, Edmundo C. Mechanics of offshore pipelines, vol. 1. London: Elsevier; 2007. [10] Palmer AC, King RA. Subsea pipeline engineering. Oklahoma: Pennwell Corporation; 2008. [11] Xue J, Hoo Fatt MS. Buckling of non-uniform long cylindrical shell subjected to external hydrostatic pressure. Eng Struct 2002;24:102734. [12] Estefen SF, Netto TA, Pasqualino IP. Strength analysis of sandwhich pipes for ultra deepwaters. J Appl Mech 2005;72:599608. [13] Showkati H, Shahandeh R. Experiments on buckling behaviour of ringstiffened pipelines under hydrostatic pressure. J Eng Mech, ASCE 2010;136(4): 46471. [14] Croll JGA. Analysis of buckle propagation in marine pipelines. J Constr Steel Res 1985;5(2):10323. [15] Chater E, Hutchinson JW. On the propagation of bulges and buckle. J Appl Mech 1984;51(2):26977. [16] ANSYS 12.1 Release. Ansys Inc. 275 technology drive. Canonsburg, PA 15317. [17] Miura K. Proposition of pseud-cylindrical concave poleyhedral shells. Report no. 442. Institute of Space and Aeronautical Science. University of Tokyo. 1969. p. 14163. [18] Miura K. PCCP shells. In: Drew HR, Pellegrino S, editors. New approaches to structural mechanics, shells and biological structures. Kluwer Academic Publishers; 2002. p. 32939. [19] DNV-OS-F101. Offshore standards. Submarine pipeline systems. Norway: Det 55 Norske Veritas; 2008.