Vous êtes sur la page 1sur 13

FINITE ELEMENT ANALYSIS OF INDENTED PIPES USING THREE-DIMENSIONAL SOLID AND SHELL ELEMENTS EXPLOITATION DELEMENTS FINIS TRI-DIMENSIONELS

SOLIDES ET CONCAVES APPLIQUES A LANALYSE NUMERIQUE DE TUBES INDENTES


T. H. Hyde, R. Luo and A. A. Becker School of Mechanical, Materials, Manufacturing Engineering and Management University of Nottingham, UK ABSTRACT This paper presents analysis of indented pipes using a number of Finite Element (FE) models to determine the deformations and stresses arising in typical cylindrical pipes damaged by indenters. The pipe material is elasto-plastic with isotropic strain hardening and the indenter is simulated with rigid contact elements with a prescribed radial displacement. Attention is focused on the residual stress distribution in the pipe wall as the indenter is removed from the pipe. A number of FE models using 2D plane strain, shell and 3D solid elements are used. A mesh sensitivity analysis and a comparison of the stresses obtained from shell and 3D solid elements show that a 3D solid model may be impractical because it requires a high degree of mesh refinement, whereas a shell FE model is capable of accurately calculating the residual stresses with considerably less computer time and storage. RESUME Cet article prsente une analyse de tubes indents, par lapplication de plusieurs modles danalyse numrique (FE) pour apprhender les dformations et les contraintes qui apparaissent dans des tubes cylindriques typiques, endommags par un indenteur. Le matriau constituant le tube est lasto-plastique avec renforcement isotrope de la rsistance et lindenteur est simul avec des lments de contact rigides qui dcrivent un mouvement prescrit vers le bas. Une attention particulire est apporte la distribution des contraintes rsiduelles dans la paroi du tube au moment o lindenteur est retir du tube. Plusieurs modles danalyse numrique utilisant des deformations planaires, des carcasses et des lments solides tri-dimensionels sont appliqus. Lanalyse de la sensibilit du rseau, ainsi que la comparaison des contraintes obtenues avec les lments carcasses et les lments tri-dimensionels, montre que le modle lments tri-dimensionels nest pas pratique parce quil exige un degr considrable de rafinement du rseau, alors que le modle danalyse numrique avec carcasse est capable de calculer les contraintes rsiduelles avec beaucoup moins de temps de calcul et de capacit de stockage dinformation.

INTRODUCTION The damage caused to buried underground pipes such as gas or oil pipelines, by an external indentation, can be detrimental to the safe operation of the pipeline. The pressure fluctuation inside the pipe can contribute to the fatigue failure of the indented pipe. Such damage is considered to be one of the main causes of failure of underground pipelines [1-4] In this paper, the ABAQUS Finite Element package [5] is used to determine the deformations and stresses arising in cylindrical pipes damaged by indenters. The indenter is simulated with rigid contact elements with a prescribed radial displacement. The material behaviour is elasto-plastic with isotropic strain hardening. The indenter is brought into contact with the pipe by applying a prescribed radial displacement. It is then removed leaving a residual stress distribution inside the pipe wall. The prescribed displacement of the rigid indenter ranges from 0.01 to 0.1 of the outer diameter (Do). Under theses circumstances, a significant amount of plasticity occurs and geometric non-linearity has to be taken into consideration. Two types of indenters are used; a long cylindrical indenter which can be analysed using a relatively simple 2D plane strain model, and a tooth-type indenter (a short cylinder with rounded ends) which requires a 3D model. Results are presented for the contact stresses and the residual stresses left within the pipe wall, due to the application of the indenter. Three FE models using 2D plane strain, shell and 3D solid elements are used. A mesh sensitivity analysis is performed to determine the number of elements needed across the thickness of the pipe. The FE solutions indicate that the stresses at the contact point of the indenter with the pipe are very sensitive to the element sizes, particularly for large indentations. The pipe models are expected to require 20 or more elements across the pipe thickness in order to produce accurate results. Therefore, a 3D model of the pipe using brick elements is not considered practical. However, since the plastic zone is localised around the indented region, it is possible to use sub-modelling to overcome this problem. The shell model uses considerably less computer storage than the 3D solid model, and is capable of producing acceptable solutions for the residual stresses, see for example [6, 7]. Based on the comparison of 3D solid and shell elements, it is shown that the shell model is valid for the analysis of the indented pipe. In this paper, details of the distribution of the residual stresses around the circumference of the pipe and along the axis of the pipe are presented for a number of indentation depths. Results based on using different FE models are compared. PROBLEM DEFINITION The geometry, boundary conditions and loading for the indented pipe are defined in Figure 1. The results presented in this paper are based on a thin-walled pipe geometry of outer diameter, Do, of 914.4 mm and wall thickness, t, of 12.7 mm (i.e. a ratio of Do/t = 72). Two types of indenters are used in this work; a long cylindrical bar indenter to model a 2D plane-strain indentation, and a 3D tooth-shaped indenter consisting of a rigid cylinder with two half-spheres at either end of the same radius as the cylinder. The material of the pipe is elasto-plastic with isotropic work-hardening with first yield at a yield stress y = 448 MPa. Figure 2 shows the stress-strain curve of the pipe material. The pipe is supported over an arc from =135o to =225o along its entire axial length, where is measured clockwise from the point of contact with the indenter. Two load steps are used to simulate an indented un-pressurised pipe. In the first load step, the indenter is brought into contact with the pipe wall by prescribing a vertical downward displacement with indentation depths ranging from 0.01 to 0.1 Do. Following each indentation depth, the indenter is removed from the pipe leaving residual stresses within the pipe wall.

indenter

Do Pipe support pipe

Figure 1: Pipe geometry, boundary conditions and loading


1000 900 800 700 600 500 400 300 200 100 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 True Strain

FINITE ELEMENT ANALYSES The ABAQUS FE package [5] is used in the analysis of the indented pipes. In all the cases, the indenter is modelled as a rigid circular surface, and is moved into the pipe by applying a prescribed displacement to the centre of the rigid surface (reference node). The FE analysis is performed with geometric non-linearity due to the large deformation of the pipe wall caused by the indenter. (a) Stress sensitivity to element sizes For a dented pipe, stress distributions at the contact point between the pipe and the indenter are complicated and sensitive to element size [8, 9]. The contact penetrating stress between the two structures can only be accurately modelled using very fine meshes. To investigate this, a quarter of 2D plane-strain dented pipe structure is analysed, as shown in Figure 3. Three different indentations of 0.05, 0.075 and 0.1 Do are applied to the un-pressurised pipe. All movements at the bottom end of the pipe are fully restrained, and the top end is constrained to move vertically due to geometrical symmetry. A number of different meshes are used where 5, 8, 15, 20, 30 and 40 quadratic 8-node elements with 2x2 integration points are placed through the wall thickness at the contact region (x=0 plane), reducing along the circumference to 5, 8, 15, 10, 15 and 20 elements, respectively, at the remote region (y=0 plane). Figure 4 shows the variation of the residual hoop stress at the contact point on the outer surface with the number of elements across the wall thickness. The FE solutions indicate that the residual hoop stresses are very sensitive to the element sizes, particularly at large indentation

True Stress (MPa)

Figure 2: Stress-strain curve of the pipe material

depths. For an indentation depth of 0.075 Do, the differences between the residual hoop stresses at the contact point vary by up to 10%, while for an indentation of 0.05 Do, they vary by up to 5%. For the larger indentation depth of 0.1 Do, the difference is about 11%.

Indenter r = 30 mm

y
4.5

t = 12.7 mm
mm

=4 R

o
Figure 3: 2D plane strain problem used for mesh refinement analysis
850 Residual hoop stress (MPa) 800 750 700 650 600 550 5 15 25 Element number 35 45
5% Do 7.5% Do 10% Do

Figure 4: Residual hoop stress at the contact point against the number of 2D quadratic elements through the wall thickness for different indentation depths To investigate the effect on the residual stresses of changing the 2D elements from quadratic 8-node to linear 4-node elements, a pipe indented to 0.1 Do is analysed with 6 different models with 5, 10, 15, 20, 30 and 60 linear, 4-node elements with reduced integration (1 integration point) through the wall thickness at the contact region. The FE solution for the contact and residual hoop stresses using the 4-node elements on both the inner and outer surfaces are shown in Figure 5 where it can be seen that contact stresses are not very sensitive to element sizes, while the residual stresses are more sensitive to mesh refinement. To compare the 4-node and 8-element solutions, Figure 6 shows the residual hoop stresses at the contact point on the outer surface obtained using both element types. This shows that FE solutions using linear and quadratic elements converge when 30 elements are placed across the thickness.

The above results indicate that using 3D solid elements would require a prohibitively large mesh which may be impractical. A shell model, however, would require less computation time than a 3D solid model, and may offer an acceptable compromise between accuracy and computational demand.
1000 800 600 Outer-residual Inner-contact 400 Outer-contact 200 Inner-residual 0 -200 5 10 15 20 25 30 35 40 45 50 55 60 -400 -600 -800 -1000
Element Number

Figure 5: Contact and residual stress at the contact plane for the inner and outer surfaces against the number of 2D linear elements through the wall thickness for an indentation depth of 0.1 Do
900 800 700 600 500 400 300 200 100 0 Residual hoop stress (Mpa)

Hoop stress (MPa)

8-node 4-node

5 10 15 20 25 30 35 40 45 50 55 60
Element number

Figure 6: Residual hoop stress at the contact point against element numbers through the wall thickness using 2D linear and quadratic elements for an indentation depth of 0.1 Do (b) FE Shell model In order to validate the shell model, an eighth of a pipe model (see Figure 7a) in which meridional movements and rotations in plane x-y on both sides CD and EF are prevented and the axial movement on side CD is restricted. All nodes on side EF are restrained to move by an equal axial distance. Displacements in the x-direction and rotations in y-z plane on side CE are prevented. On side DF, displacements in the x and y directions and all rotations are prevented. FE analyses are performed for different indenter radii ranging from 12.7 to 40 mm, internal pressures ranging from 0 to 5.12 MPa, indentation depths ranging from 0.02 Do to 0.05 Do, and wall thicknesses ranging from 12.7 to 27 mm. The FE shell model consists of 10845 linear, 3-node and 4-node reduced integration (with 1 integration point) finite-strain elements. Two 2D models are used for comparison purposes,

one consisting of 9525 linear 3-node and 4-node elements with 1 integration point, and the other consisting of 3656 quadratic 6-node and 8-node elements with 3 and 4 integration points, respectively. The 2D linear and quadratic element models contain 30 and 32 elements, respectively, across the wall thickness at the contact point. Part of the shell mesh is shown in Figure 7b.
A
E

A C G y z

B C
F

x O

(a) Geometry

(b) FE shell element mesh of region ABGC

Figure 7: Shell model and FE shell element mesh used to model plane strain conditions Different indenter sizes. In order to compare the 2D and shell models for different indenter sizes, different indenter radii of 12.7, 20, 30 and 40 mm are used. An internal pressure of 3.84 MPa is applied before indentation. The indentation depth is 0.035 Do and the pipe wall thickness is 12.7 mm. The contact and residual hoop stresses at the contact position on the outer and inner surfaces, calculated by using shell elements and linear 3-node and 4-node 2D elements, are shown in Table 1. The results indicate that differences in the contact stresses between shell and 2D models are very large, up to 25 %, whereas the differences between the residual hoop stresses are much smaller at less than 5.2 %. This can be explained by the fact that the shell model cannot accurately simulate the variation of the penetrating pressure across the wall thickness caused by the indenter. A comparison of the residual stresses obtained by using shell and quadratic 2D elements is shown in Table 2, where it can seen that the differences are less than 5%. Table 1: Contact hoop stress solutions for the shell and 2D linear element models for different indenter radii at the contact position
Indenter radius (mm) On outer surface Shell (MPa) 2D (MPa) Error (%) Shell (MPa) On inner surface 2D (MPa) Error (%)

12.7 20 30 40

-749 -744 -743 -743

-999 -993 -969 -895

25 25 23 17

789 784 783 783

740 740 740 738

6.6 5.9 5.8 6

Table 2: Residual hoop stress solutions for the shell and 2D models for different indenter radii at the contact position
On outer surface Indenter radius (mm) Shell 2D linear Difference 2D Difference Shell (MPa) (MPa) (%) quadratic (%) (MPa) On inner surface 2D Difference 2D Difference linear (%) quadratic (%) (MPa)

12.7 20 30 40

741 734 729 733

704 702 701 702

5.2 4.5 3.9 4.4

737 690 697 700

0.4 6.4 4.6 4.7

-773 -766 -768 -764

-736 -736 -735 -734

5 4.1 4.5 4.1

-744 -743 -741 -741

3.9 3.1 3.6 3.1

Different internal pressures. To investigate the influence of the internal pressure on the shell and 2D solutions, the FE simulations are performed by applying different internal pressures of 0, 2.56, 3.84 and 5.12 MPa before indentation. An indenter of radius 30 mm is used with an indentation depth of 0.035 Do and pipe wall thickness of 12.7 mm. The residual stresses are listed in Table 3, which show that changing the internal pressure causes differences of up to 5.9% between the shell and 2D models. Table 3: Residual hoop stress solutions for the shell and 2D models for different pressures at the contact position
On outer surface Pressure (MPa) Shell 2D linear Difference 2D Difference Shell (MPa) (MPa) (%) quadratic (%) (MPa) On inner surface 2D Difference 2D Difference linear (%) quadratic (%) (MPa)

0 2.56 3.84 5.12

560 716 729 741

544 682 701 711

2.9 4.9 4 4.2

553 688 697 710

1.2 4.1 4.6 4.3

-532 -734 -768 -789

-502 -708 -735 -752

5.9 3.7 4.5 4.9

-539 -715 -741 -758

1.3 2.6 3.6 4.1

Different indentation depths. Indentation depth is the main factor that causes a dented pipe to fail either during indentation or afterwards due to fatigue failure caused by pressure fluctuations. Indentation depths of 0.02, 0.035 and 0.05 Do are applied in order to investigate the solution differences between the shell and 2D models. An internal pressure of 3.84 MPa is applied before indentation. An indenter of radius 30 mm is used with a wall thickness of 12.7 mm. The residual hoop stress solutions are shown in Table 4, where it can be seen that changing the indentation depth can produce differences of up to 8.6% between the shell and 2D solutions for the range of conditions investigated. Table 4: Residual hoop stress solutions for the shell and 2D models for different indentation depths at the contact position
On outer surface Indenter depth Shell 2D linear Difference 2D Difference Shell (MPa) (MPa) (%) quadratic (%) (MPa) On inner surface 2D Difference 2D Difference linear (%) quadratic (%) (MPa)

0.2 Do 0.35 Do 0.05 Do

629 729 795

579 701 760

8.6 4 4.6

648 697 783

2.9 4.6 1.5

-652 -768 -826

-626 -735 -784

4.2 4.5 5.3

-637 -741 -791

2.3 3.6 4.4

Different wall thicknesses. Increasing the pipe wall-thickness can improve resistance to external impact loading and thus improve the safety of the pipe. In this section, different wall thicknesses of 12.7, 17, 22 and 27 mm are used to compare the shell and 2D solutions without changing the number of elements across the wall thickness. An indenter of radius 30 mm is used with an indentation depth of 0.035 Do and an internal pressure of 3.84 MPa. A comparison of the residual stress solutions between the shell solutions and the 2D linear and quadratic elements are listed in Table 5. The results show that at the contact point on the outer surface, residual stress differences between the shell and 2D models increase as the wall-thickness increases. Table 5: Residual stress solutions for the shell and 2D models for different wall thicknesses at the contact position
On outer surface Wallthickness (mm) Shell 2D linear Difference 2D Difference Shell (MPa) (MPa) (%) quadratic (%) (MPa) On inner surface 2D Difference 2D Difference linear (%) quadratic (%) (MPa)

12.7 17 22 27

729 743 750 757

701 703 703 700

4 5.6 6.7 8.1

697 718 703 705

4.6 3.5 6.7 7.4

-768 -760 -757 -758

-735 -733 -726 -712

4.5 3.6 4.3 6.5

-741 -737 -730 -725

3.6 3.1 3.7 4.5

The above results show that stress solutions are very sensitive to mesh refinement across the pipe wall, and can only be obtained using 30 or more elements across the wall thickness for a Do/t ratio of 72. This implies that in a 3D dented pipe FE model, the number of elements needed to obtain accurate stress solutions will be very large. The above results also show that the shell FE models may not be capable of accurately calculating the contact stresses but are capable of calculating residual stresses to a reasonable degree of accuracy. Although 2D quadratic elements are considered to be most capable of accurately calculating stress values, the solutions listed in the above tables indicate that for most cases using 2D linear elements can give residual stresses on both surfaces very close to those of the quadratic elements. In the analyses of the effects of the indenter radius, the indentation depth, the internal pressure and the wall thickness, on the differences in the solutions between the shell and 2D models, the FE solutions show that these factors do not significantly change the residual hoop stress between the shell and 2D models. (c) 3D solid models To model a tooth-like indenter, rather than a long cylindrical indenter, a shell or 3D solid model is needed. In this section, a short rigid cylindrical indenter is applied with a radius of 12.7 mm and an axial length of 20 mm with two half-spheres at either end with the same radius as the cylinder. An un-pressurised pipe in indented to a depth of 0.01 Do and is then released. A 3D solid FE model of the dent region is created as a sub-model rather than a full model in order to increase the mesh refinement at the contact region. The sub-model is generated with 9 elements through the wall thickness in the dent region and 3 elements across the thickness elsewhere, as shown in Figure 8. Three types of quadratic solid elements are used; 20-node and 27-node brick elements and 15-node triangular prism elements with reduced integration. The shell FE model, shown in Figure 9, is generated with 4-node doubly-curved quadrilateral elements and 3-node triangular elements with reduced integration, where fine meshes are created in the dent region and coarse meshes elsewhere. The FE solutions for the variation of hoop, axial and von Mises stresses obtained using 3D solid and shell models are shown in Figures 10 and 11 along the circumference at the axial

plane z = 0, and Figure 12 along the axial distance from the point of contact at the = 0o plane. From these figures, it can be seen that the residual stress distributions on the pipe outer surface obtained using the 3-D solid model are not smooth but oscillatory in the dent region especially for the hoop and axial stress distributions. This phenomenon occurs because the meshes used for the 3-D solid analysis are not refined enough, even when 9 elements are placed across the wall thickness. Comparing the residual stress distributions obtained using the shell and 3D solid models, it is clear that they are quite consistent on the pipe inner surface, except at the dent region on the pipe outer surface where oscillations of hoop and axial stress occur in the 3D solid model.. However, von Mises equivalent stress distributions in the 3-D solid model are not oscillatory and are close to those in the shell model. Hence, the comparisons of the solutions obtained employing the shell and 3D solid models show that the shell model can adequately simulate the residual stresses resulting from pipe indentation.
FEMGV 5.1-03 Model: A1 Nottingham University (SG) 11 OCT 1999

Y X Z

Figure 8: Sub-model of the 3D solid element model


Analysis: NEUTRAL

X Y Z

Figure 9: Shell mesh used to model 3D conditions

Residual stress (MPa) on inner surface

800 600 400 200 0 -200 -400 -600 -800 Circumference angular position() 0 10 Hoop stress 20 30 40 Axial stress von Mises

(a) Shell model 800


Residual streses (MPa) on inner surface

600 von Mises 400 200 0 -200 -400 -600 -800 Circumferential angular position() 0 Axial stress 10 20 Hoop stress 30 40

(b) 3D solid model Figure 10: Residual stress along the inner circumference of the pipe for an indentation depth of 0.01 Do at the z = 0 plane

Residual streses(MPa) on outer surface

800 700 600 500 400 300 200 100 0 -100 0 10 20 30 40 Circumferential angular position () Axial stress Hoop stress von Mises

(a) Shell model

900 Rresidual stress (MPa) on outer surface 800 700 600 500 400 300 200 100 0 -100 0 10 20 30 40 Circumferencial angular position () Axial stress Hoop stress

von Mises

(b) 3D solid model Figure 11: Residual hoop stress along the outer circumference of the pipe for an indentation depth of 0.01 Do at the z = 0 plane

900

Residual stress (MPa) on outer surface

800 700 600 500 400 300 200 100 0 -100 -200 0 50

Hoop stress von Mises

Axial stress

100

150

200

250

300

350

Axial position (mm) (a) Shell model Re sid ua l str es s (M pa ) on ou ter su rfa 800 700 600 500 400 300 200 100 0 -100 0 -200 Axial position (mm) (b) 3D solid model Figure 12: Residual hoop stress along the outer axial position the pipe for an indentation depth of 0.01 Do at the = 0o plane 50 100 150 200 250 300 350 Axial stress von Mises Hoop stress

CONCLUSIONS Due to the localised plastic deformation of indented pipes, FE solutions are quite sensitive to mesh refinement in the dent region, requiring more than 30 elements across the wall thickness for accurate results to be obtained. This stress sensitivity to element sizes demonstrates that to obtain accurate stresses in the dent region using a 3D solid element model would require an excessive number of elements. Comparisons of the FE solutions obtained using shell elements with those obtained from 2D and 3D solid elements for different indenter radii, internal pressures, indentation depths and wall thicknesses, show that shell models are capable of accurately calculating the residual stresses in dented pipes. ACKNOWLEDGEMENTS The authors wish to acknowledge the financial support and technical direction of Advantica and the University of Nottingham for this work through a University HUB scholarship for Mr. Luo. The opinions of the authors expressed in this paper do not necessarily represent those of Advantica. REFERENCES 1. Shannon, R.W.E. (1974). The failure behaviour of line pipe defects, Int. J. Pressure Vessels & Piping, 2, 243-255. 2. Fearnhough, G.D. and Jones, D.G. (1978), An approach to defect tolerance in pipelines, Conference on Tolerance of Flaws in Pressurised Components, I. Mech.E., London. 3. Ainsworth, R.A. (1984). The assessment of defects in structures of strain hardening material, Eng. Fracture Mechanics, 19, 633-642. 4. Bai, Y. and Song, R.(1998). Fracture assessment of dented pipes with cracks and reliability based calibration of safety factor, Int. J. Pressure Vessels & Piping, 74, 221-229. 5. ABAQUS (1998). Version 5.5, Hibbitt, Karlsson and Sorenson, Inc. 6. Alexander C. R. (1999). Analysis of dented pipes considering constrained and unconstrained dent configurations, Energy Sources and Technology conference & Exhibition, ASME, ETCE99-6686. 7. Fowler, J. R., Alexander, C.R., Kovach, P.J. and Connelly, L.M. (1995). Fatigue life of pipes with dents and gouges subjected to cyclic internal pressure, American Society of Mechanical Engineers, Petroleum Division, PD, 69, 17-35. 8. Godoy, L.A. (1990). A perturbation formulation for sensitivity and imperfection analysis of thin-walled structures, Latin American Applied Research, 20, 147-153. 9. Leis, B.N, Francini, R.B., Mohan, R., Rudland, D.L. and Olson, R.J.(1998). Pressuredisplacement behaviour of transmission pipes under outside forces-towards a serviceability criterion for mechanical damage, Proceedings of the International Offshore and Polar Engineering Conference, 2, 60-67.

Vous aimerez peut-être aussi