Sie sind auf Seite 1von 8

Applied Ocean Research 30 (2008) 297304

Contents lists available at ScienceDirect

Applied Ocean Research


journal homepage: www.elsevier.com/locate/apor

Analysis of testing methods of pipelines for limit state design


F. Guarracino , M. Fraldi, A. Giordano
Dipartimento di Ingegneria Strutturale, Universit di Napoli Federico II, 80125 Napoli, Italy

article

info

abstract
It is well known that the design of submarine pipelines relies on accurate test results for the local buckling collapse of pipes subjected to bending loading. The present paper analyses apparently anomalous values of axial tensile and compressive strains from recent test results in comparison to the values that would be expected on the basis of simple bending theory. This could have important consequences for the efficacy of the design factors derived using these results. The cause of the differences between the strain values obtained in the tests and those expected on the basis of simple bending theory are explained using finite element modelling. The differences result from the type of collars and supports commonly used in bending tests, the effects of which persist for a greater length along the test pipe than has hitherto been assumed. In general, it is pointed out that the application of the simplified engineering theory of bending can be erroneous when ovalisation is imposed or, on the contrary, the boundary conditions of the section are restrained from ovalising deformations. The influence of the D/t ratio is also analysed. The results contribute to the understanding of a crucial limit state for the design of onshore and offshore pipelines. 2009 Elsevier Ltd. All rights reserved.

Article history: Received 19 December 2007 Received in revised form 21 December 2008 Accepted 26 December 2008 Available online 4 February 2009 Keywords: Pipelines Testing arrangement Limit state

1. Introduction In most modern pipeline designs, the required minimum wall thickness is determined on the basis of a maximum allowable hoop stress under design pressure. Generally, the initial wall thickness design is based on the assumption that pressure will be the governing load. However, a pipeline may be subjected to additional loads due to installation, seabed contours, impacts and high-pressure/high-temperature operating conditions for which the bending moment capacity is often the limiting parameter [1]. The design calculations for pipelines are aimed at providing a safe, robust pipeline with an economical use of expensive material and installation equipment. Pipeline design has traditionally been based on a limiting stress approach but since 1996 a limit state code has been developed by DNV with the latest version issued in 2000 [2]. The use of the limit state approach provides a more comprehensive basis for the calculation of the ultimate conditions for pipes subjected simultaneously to pressure and bending loads. One aspect that is incorporated in the code is the concept of ultimate states of pipe with load-controlled and displacementcontrolled loading. In the former, the bending deformations are directly affected by the variation of the applied moments. The ultimate state thus relates to a maximum moment condition.

Corresponding address: Dipartimento di Ingegneria Strutturale, Universit di Napoli Federico II, via Claudio, 21, 80125 Napoli, Italy. Tel.: +39 081 7683733; fax: +39 081 7683233. E-mail address: fguarrac@unina.it (F. Guarracino).
0141-1187/$ see front matter 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.apor.2008.12.003

In the displacement controlled condition the deformations are controlled, say by bending the pipe round a prescribed surface, and once the pipe is in contact with the surface, additional loading of whatever kind, could not increase the local curvature of the pipe. This form of loading is associated with a maximum curvature, i.e. a maximum local axial strain. In both these cases, the ultimate state of the pipeline deformation or loading is calculated using a model that describes the characteristic ultimate moment or strain related to the geometry and material properties of the pipe. The characteristic values are then modified using design factors that produce the design values of allowable ultimate moment or strain. The models used to determine the characteristic levels of moment or strain are generally chosen as those describing as closely as possible mean values of these variables measured during tests. The design factors are calculated using statistical descriptions of the scatter of test results compared to the mean values together with the statistical descriptions of the variables composing the particular model, e.g. material strength, modulus etc. The calculation to determine the design factors also include a calibration stage in which the values of the design factors are adjusted to ensure that the results from applying the code formulation and factors prescribe levels of the ultimate conditions that conform to industry-wide acceptable levels for the probability of failure of the pipe. In the process described above, it is generally assumed that the scatter of test results arises from minor and usually random variations in the variables included in the design model. In the case of a pipe, these variations would generally relate to differences in material properties and in the geometries of the test pipes (such

298

F. Guarracino et al. / Applied Ocean Research 30 (2008) 297304

Fig. 1. Typical four-point bending test arrangement.

as wall thickness or out-of-roundness) from their corresponding nominal values. The design model is intended to provide a description of the test conditions that account for systematic differences between one test and another. In the present paper the apparently anomalous results from tests on pipes, which have been observed, but ignored, in the past, are analysed and explained with the aid of finite element simulations. The influence of the D/t ratio is also examined and the potential influence that these anomalies might have on the process of providing design calculation guidance using the limit state method is highlighted. 2. Pipe testing The pipe cross sectional bending moment is directly proportional to the pipe curvature. The moment curvature relationship provides information necessary for design against failure due to bending. If the pipe is part of a carrying structure, the elastic limit may be an obvious choice as the design limit. However, for pipelines and risers where the global shape is less important, this criterion will be overly conservative due to the significant resources in the elasticplastic range. Higher design strength can therefore be obtained by using design criteria based on the stress/strain levels reached at the point of onset for local buckling or at the ultimate moment capacity. For displacement-controlled configurations, it can even be acceptable to allow the deformation of the pipe to continue into the softening region. Results from testing a section of a pipe in purely bending loading has usually been analysed on the basis that simple bending theory can adequately describe the behaviour of the test specimen. Primarily this implies that while the pipe material remains elastic the application of purely bending moment will induce maximum tensile and compressive strains that are identical in magnitude. A typical test rig for a medium diameter pipe, of about 700 mm diameter, is shown in Fig. 1. The test rig applies a four-point bending condition with the central section of the test pipe assumed to be subjected to bending action only, with no, or at most very little, shear or axial forces. The loading of the section of pipe during a test has slightly different requirements from that pertaining to a section of a very long continuous pipe. At the four loading points, see Fig. 1, the pipe is reinforced to avoid the external loads from locally deforming and perhaps crushing the pipe. Since the pipe is assumed to be a simple structural element so that simple beam theory holds true, it has been common practice to assume that the axial strains have identical values in tension (top surface) and compression (bottom surface) and that the strains can be calculated directly from the curvature or the vertical displacements of the central section of the pipe. The ultimate strain values from tests in which the pipe has been loaded to the point of local buckling have usually been inferred from measurements of the deformations. Only recently have strain gauges been attached to the test pipe to measure axial strains directly. Some time ago tests [3] were carried out on 152 mm diameter pipes to determine the maximum curvature to which

Fig. 2. Results from a 152 mm diameter pipe bend test [3].

the pipe could be deformed prior to local buckling occurring. An arrangement similar to that in Fig. 1 was used, although very thick steel collars were used to protect the pipe at the loading points. The collars were machined to fit very closely around the pipe to ensure no localised loading was applied to the pipe wall. As a result, the pipe was fully prevented from ovalising at the loading points. Strain gauges to measure axial and circumferential strains were attached at intervals of 100 mm apart along the central test section. In the design of the test rig it was assumed that a central test section of about 5D would suffice to ensure that end effects due to the loading conditions would diminish to a negligible level along the major part of that section. Fig. 2 shows results of the axial strain values along the top and bottom of the pipe section for two levels of the applied loading. It is evident the axial strains are fairly uniform along the length of the test section but that there are significant differences in the averaged values of compressive and tensile strains. At the lower level of applied loading, the averaged compressive axial strains are seen to be about 1.36 times the corresponding values of the tensile strains. The difference between the averaged axial strain levels increased for the higher level of loading with the averaged compressive strains being 1.59 times the corresponding tensile strains. At that time the evident anomaly between the measured strains with the expected values vis--vis the simple bending theory was not followed up, and even after checking that the strain gauges were correctly positioned and the instrumentation was functioning properly the cause of the anomaly was not further investigated. Proving tests were later carried out on several sections of 609 mm diameter pipes containing a thin liner made from a corrosion resistant material [4]. The purpose of the tests was to determine accurately the level of strain to which the pipe could be bent before the liner buckled locally. The test arrangement is shown in Fig. 3 in which the loading arm is 2 m long to create the moment in the central section of the test pipe. The test section was

F. Guarracino et al. / Applied Ocean Research 30 (2008) 297304

299

Fig. 6. Test arrangement with modified support and load application points. Fig. 3. Test arrangement for lined pipe proving test [4].

Fig. 4. Averaged strain values plotted against corresponding values of applied loading [4].

3.5D long. It may be seen from Fig. 3 that the load is applied to the test pipe using straight bars and loose yokes around part of the pipe circumference. A number of axial strain gauges were attached along the top and bottom centre lines of the pipe at intervals from the support points. The values of strain were monitored as the load values were progressively increased and were found to be uniform along the test section. Figs. 4 and 5 show plots of the variation of the values of the top and bottom strain gauges averaged along the test sections against corresponding values of applied load for two pipes of different wall thickness, respectively. It is evident from Figs. 4 and 5 that in both cases there is a systematic difference between averaged compressive strains compared to corresponding tensile strains. In one test, Fig. 4, the averaged axial tensile strains were 1.21 times the corresponding compressive strains and in the other test, Fig. 5, the averaged axial tensile strains were 1.28 times the corresponding averaged compressive strains. In view of the importance of the results of the tests in providing the allowable levels of strain for the lined pipe an investigation was made with regard to the underlying cause of the anomaly. This is described in the next section of this paper and some preliminary studies can be found in [5,6]. The investigation determined that the cause lay in the effect of the imposed ovalisation applied by the saddles at the load points. This result pointed to a proposal for the modification of the loading application in which the loads were applied, not through local stiffening of the pipe wall or saddles, but through the neutral axis of the pipe, as shown in Fig. 6. The test pipe was fitted with strain gauges, as before, and also gauges to measure the ovality of the pipe. As before, the values of the axial strains measured by the gauges along the test section of the pipe were very uniform. Fig. 7 shows the result from averaging the measured values of strain at positions along the test section plotted against corresponding values of applied load. It may be seen that with the modified loading and support arrangement, the averaged measured values of compressive strains agree very closely with the corresponding values of the tensile strains. 3. Analysis of the effects of testing arrangement Following the observation of the apparent anomaly in the variation of the tensile and compressive strains compared with the values expected on the basis of the simple theory of bending, analytical [7] and numerical modelling has been carried out to investigate the root cause of the anomaly. Essentially, it was found that during the test of a short section of pipe, the practical loading and support arrangements can result in boundary conditions that may impose some degree of ovalisation at the point of load application or, alternatively, decrease the development of the natural ovalisation [8,9]. It has generally

Fig. 5. Averaged strain values plotted against corresponding values of applied loading [4].

300

F. Guarracino et al. / Applied Ocean Research 30 (2008) 297304

Fig. 9. Finite element simulation of a test arranged as in Figs. 1 and 3: strain values at the mid-section of the pipe applied load 1 MN.

Fig. 7. Averaged measured values of strain at positions along the test section with loads applied through the neutral axis of the pipe.

Fig. 8. Outline of the finite element modelling of a test arranged as in Figs. 1 and 3.

been assumed that such boundary effects would have a minor consequence on the deformations of the test pipe and would persist for only a short distance along the test length. However, the investigation has shown a hitherto unsuspected mechanism in which the imposition or the prevention of the ovalisation at the loading point of the test pipe will set up an axial strain system that is additional to the usual axial strain caused by simple bending. The conjunction of the two strain systems thus causes a difference between the values of the axial compressive and tensile values. The interaction between the ovalisation and the axial strain effects is examined here using finite element modelling. Fig. 8 shows the finite element simulation of a test arranged as in Figs. 1 and 3 on a 609.6 mm diameter pipe (t = 18.9 mm). For this purpose the commercially available nonlinear finite element code, ANSYS r v.11.0, was employed. The case under consideration was modeled by means of 9000 four-node SHELL63 elements for an overall length of 8 m. According to ANSYS documentation [10], SHELL63 has both bending and membrane capabilities. Both in-plane and normal loads are permitted. The element has six degrees of freedom at each node: translations in the nodal x, y, and z directions and rotations about the nodal x,

y, and z axes. Stress stiffening and large deflection capabilities are included. A consistent tangent stiffness matrix option is available for use in large deflection (finite rotation) analyses. Both the loading and the support zones were supposed to span over a length of one fourth of the mean radius of the pipe. In this respect a preliminary mesh refinement sensitivity analysis was carried out. The material properties of a generic isotropic high-grade steel, i.e. E = 2.07 105 N/mm2 and = 0.3, were assumed. Both linear and nonlinear elastic analyses were performed and it was found that a linear analysis was sufficient to deliver very accurate results. The FE strain distribution along the mid-section of the pipe for an applied load of 1 MN is shown in Fig. 9 for elastic conditions and is in good agreement with the experimental results of Fig. 5. It is evident that the tensile strains at the top of the pipe are about 1.25 times the compressive strains at the bottom of the pipe. As a matter of fact, the rigid loading yokes at the supports and at the loading points induce some degree of ovalisation on the pipe and this results in an increased value of the tensile strains at the top of the pipe, as confirmed by the finite element analysis. This outcome makes it clear that in test design and result analysis special attention must be paid to the effects of loading and constraint arrangement of the pipe being tested. Fig. 10 shows the values along the pipe axis of the longitudinal strains at the top, at the bottom and at the side of the pipe, as yielded by the finite element analysis. It may be seen that the effects of the loading and of the constraints propagate in quite a complex manner along the full length. It is also evident in the figure that at the mid-span of the model, where the bending moment can be considered constant, the ratio of the tensile axial strain to the corresponding compressive strain is about 1.25 times. Since generally this is the section of pipe that is assumed to be free from the boundary support effects and to have strain levels pertaining to simple bending theory, it can be seen that the analysis confirms the anomaly observed in the tests, see Figs. 4 and 5, and also confirms that the presence of the test loading conditions will affect the axial strain levels at which local axial buckling will be initiated in a test pipe. Finite element modelling has as well been applied to determine the length over which the effects of the loading and of the boundary constraints propagate and hence influence the axial strains. In order to do so, a very simple modelling was applied to pipes with different diameters and wall thicknesses. The modelling consisted in applying a vertical load of 1 MN exactly on a vertical support. In this manner no bending moment was expected along the axis of the pipe and all the strains were due to the imposed ovalisation of the loaded section. As seen before, these strains are essentially similar to those which, in the case of bending tests, add to the bending

F. Guarracino et al. / Applied Ocean Research 30 (2008) 297304

301

Fig. 10. Finite element simulation of a test arranged as in Figs. 1 and 3: strain values along the pipe axis applied load 1 MN.

Fig. 11. FEM simulation for a pipe with D/t = 32 (diameter 609.6 mm, wall thickness 18.9 mm): strain values along the pipe axis applied load 1 MN.

strains and can cause a significant alteration in the symmetry with respect to the neutral axis of the section. It is worth highlighting that this effect cannot be seen as a classic local perturbation, as is shown in the next figures. Three D/t ratios were taken into consideration, i.e. 16, 32 and 64, obtained by varying either the diameter or the wall thickness of the pipes. The results are shown in Figs. 1115. On account of

the symmetry, in all the plots the strain values at the top and at the bottom of the pipe coincide. First, from Figs. 1115 it is evident that the maximum and the average value of the axial strain cannot be directly related to the D/t ratio. In fact, when the D/t ratio increases on account of a reduction in wall thickness (the diameter D is kept constant at 608.6 mm, see Figs. 12, 11 and 13 in sequence), the intensity of the

302

F. Guarracino et al. / Applied Ocean Research 30 (2008) 297304

Fig. 12. FEM simulation for a pipe with D/t = 16 (diameter 609.6 mm, wall thickness 37.8 mm): strain values along the pipe axis applied load 1 MN.

Fig. 13. FEM simulation for a pipe with D/t = 64 (diameter 609.6 mm, wall thickness 9.45 mm): strain values along the pipe axis applied load 1 MN.

strain tends to become greater with the D/t ratio. On the contrary, when the D/t ratio increases on account of a larger diameter (the wall thickness is kept constant at 18.9 mm, see Figs. 14, 11 and 15 in sequence), the intensity of the strain tends to reduce. Second, it can be noticed that, conversely, the natural wavelength of the solution can be directly related to the D/t ratio. In fact, from Figs. 1115 it can be seen that when the D/t ratio increases (either on account of a reduction in wall thickness or an

increase in diameter), the effects of the loading propagate with a wavelength which tends to become greater with the D/t ratio. Largely, it can be affirmed that the effects of loading and constraint arrangement persist for a greater length along the test pipe than has hitherto been assumed. Equally, when the ovalisation of the cross section under bending (which takes place on account of the well-known von Krmn effect [8,9]) is restricted by any form of constraint, a marked

F. Guarracino et al. / Applied Ocean Research 30 (2008) 297304

303

Fig. 14. FEM simulation for a pipe with D/t = 16 (diameter 304.8 mm, wall thickness 18.9 mm): strain values along the pipe axis applied load 1 MN.

Fig. 15. FEM simulation for a pipe with D/t = 64 (diameter 1219.2 mm, wall thickness 18.9 mm): strain values along the pipe axis applied load 1 MN.

asymmetry between the compressive and tensile strains occurs with a turnround with respect to what has been shown in the previous finite element analyses. This is the case of the results from a 152 mm diameter pipe bend test [3], shown in Fig. 2, where the averaged compressive axial strains are seen to be about 1.36 times the corresponding values of the tensile strains. The explanation of this phenomenon is quite simple: preventing the natural ovalisation under bending can be seen as applying a loading similar to those in Figs. 1115, but reversed in sign. The resulting

strains will be reversed in sign, too, and, in the case of Fig. 3, add to the bending strains, causing an increase in the value of the compressive strains and a reduction in the value of the tensile strains. 4. Considerations on the limit state design As said earlier, depending on the function of the pipe, any of the points from the moment curvature relationship can be used as the

304

F. Guarracino et al. / Applied Ocean Research 30 (2008) 297304

decreasing the apparent buckling strain in a test it is possible that the calculations using the apparent values of buckling strain will result in factors that do not actually correspond to the intended level of probability of failure. 5. Conclusions The focus of the present work has been to identify and analyse the effects of the test arrangement on the level of apparent strain, based on the assumption of simple engineering bending theory, at which local buckling is initiated. The apparently anomalous values of measured axial strain in tests have been explained with the aid of finite element modelling. The length over which the effects of the loading and of boundary constraints propagate (and hence influence the axial strains) has been investigated. Hence it has become evident that the scatter in test results from different types of testing arrangement could include a systematic error that is induced by the constraints at the support and loading points in the test rig. The results contribute to the understanding of a crucial limit state for the design of onshore and offshore pipelines. Acknowledgements The authors wish to express their gratitude to Prof. A.C. Walker for providing results and information from the tests described in [3]. The financial support from ReLUIS (University Network of Seismic Engineering Laboratories) and from the International Mobility Program of the University of Naples Federico II is also graciously acknowledged from the first author. References
[1] Guarracino F, Mallardo V. A refined analytical analysis of submerged pipelines in seabed laying. Applied Ocean Research 1999;21:28193. [2] Design standard OS F101 offshore pipelines. Det Norske Veritas; 2000. [3] Ellinas CP, Walker AC, Langfield GN, Vines MJ. A development in the reeling method for laying subsea pipeline. In: Proc. 1st petroleum technology Australian conference. 1985. [4] Walker AC, Holt A, Guarracino F, Wilmot D. Test procedure for pipe and pipeline material. In: Offshore pipeline technology conference. 2003. [5] Guarracino F. On the analysis of cylindrical tubes under flexure: Theoretical formulations, experimental data and finite element analysis. Thin Walled Structures 2003;41:12747. [6] Guarracino F, Giordano A. Influence of boundary conditions on the instability of circular cylindrical shells. In: Proceedings of ASSCCA03. 2003. [7] Guarracino F. Analytical evaluation of local effects in cylindrical shells testing and design. Strength of Materials 2008 [in press]. [8] von Krmn Th. Ueber die Formnderung dnnwandinger Rohre, insbesondere federnder Ausgleichrohre. Zeitschrift des Vereines deutscher Ingenieur 1911;45:188995. [9] Brazier LG. On the flexure of thin cylindrical shells and other thin sections. Proceedings of the Royal Society A 1927;116:10414. [10] ANSYS. Inc. ANSYS 11.0 users documentation. Canonburg, PA 15317. USA; 2007. [11] Gresnigt AM, van Foeken RJ. Local buckling of UOE and seamless steel pipes. Paper No. AMG-2001-04. In: ISOPE conference. 2001.

Fig. 16. Scatter of experimental strain versus D/t from various reported tests in the literature [11].

design limit. In certain cases the elastic limit may be an obvious choice but for pipelines and risers this criterion will be excessively conservative due to the significant resources in the elasticplastic range. Higher design strength can therefore be obtained by using design criteria based on the stress/strain levels reached at the point of onset for local buckling or at the ultimate moment capacity. This can be justified provided the availability of the carrying capacity with high deformations combined with a precise prediction of the deformation pattern and its amplitude. The results from bending tests have shown that depending on the form of arrangement at the support and at the point where the load is applied to the pipe, the compressive strains might be greater or less than the corresponding tensile strains, by up to 30%. It is also recognised that the onset of local buckling is instigated by the level of compressive axial strain in the pipe wall. Thus, the variation in the axial compressive strain due to the loading arrangement, i.e. the degree of constraint on the ovalisation at the pipe at that position, will affect the level of overall bending that would initiate local buckling. In light of this, it is not unreasonable to expect that a comparison of the values of strains at which buckling has been initiated in tests carried out by various researchers, with different test arrangement, will show a degree of scatter. This scatter has tended to be assumed to have resulted primarily from variations in the geometry of the test pipes, usually in the form of out-ofroundness of the pipe wall. However, as a result of the analysis of effects due to the test loading conditions, it is now suspected that a significant component of the scatter could result from differences in the testing equipment. Fig. 16 shows a few examples from various reported tests in the literature [11]. These results, together with others, have been used in the calculations of the appropriate values of design factors that are currently used in the limit state design of subsea pipelines. In view of the effects of the test arrangement in increasing or