corrosion assessment

7/18/2019 Corrosion assessment

http://slidepdf.com/reader/full/corrosion-assessment 1/6

Development of an Ansys Interface for FE Solid Modeling and Analysis

of Corroded Pipelines

RITA C. C. SILVA, JOÃO N. C. GUERREIRO and PATRÍCIA R. C. DRACH National Laboratory for Scientific Computing (LNCC)

Av. Getúlio Vargas, 333, Petrópolis – RJBRAZIL

Abstract: - The assessment of the structural integrity of corroded pipes has become of main interest to helpengineers to take the important decision of maintaining or repairing a pipeline. Solid finite element (FE) modelshave been widely used to perform failure analyses but the generation of the required models is, in many cases, ahard task to do. In this work, we present the PIPE program that provides an Ansys Interface to automaticallygenerate solid models and to manage the FE analyses of pipes with multiple rectangular defects in arbitrary position. The code is totally developed within the general-purpose program Ansys and the whole process is

guided by a friendly interface. A validation test comparing the experimental to the numerical results is presented.

Key-Words: - Corroded Pipeline, FE Modeling, Mesh Generator, Rupture Analysis, Numerical Simulation

1 IntroductionThe metal loss due to corrosion defects has becomeone of the leading causes of pipeline failure. Areliable assessment of the remaining strength of pipescontaining single or multiple corrosion defects is acontinuous matter of interest to engineers in order toreduce economical and environmental cost. Numerical approaches as the Finite Element Method

(FEM) can be used to evaluate the pipeline integrityand to figure out the interaction effects due tomultiple corrosion defects, as performed in [1,2,3].FE non-linear analysis of solid models requiresexpertise during all the processing stages, which mayspend a great deal of time in computation and humaneffort.

In this work, we present a program, named PIPE,which we developed to, automatically, generate andsolve solid models of pipes in the Ansys background,which is a worldwide accepted commercial FE code.The PIPE program allows a fast model generation ofundamaged pipes and of corroded pipes with singleor multiple rectangular defects in arbitrary position.Defect patch configurations with double,longitudinal, circumferential and no symmetries can be modeled.

The undamaged pipe models are designed withuniform meshes and the corroded pipe models, with aminimum of 2 and a maximum of 5 mesh refinementzones. The transitions between adjacent refinedregions involve a mesh transition along the pipe walland a mesh transition along the pipe surface, doubling

the element length size.

The rupture analysis is performed automatically by the restart command of the Ansys program,adopting a controlled time step and a failure criterionused in joint projects of LNCC and Petrobras R&DCenter.

Numerical and experimental results are comparedto demonstrate the program effectiveness.

2 The Pipe ProgramThe Pipe program was conceived to produce anAnsys interface to quickly generate 3D models and to perform elastic-plastic geometric non-linear analysesof non-corroded and corroded pipes.

The program is composed by a set of subroutines,containing standard Ansys commands and Ansys parametric design language (APDL) commands [4].When executing an input file, Ansys is normallyrestricted to a linear flow and each statement isexecuted in the order that it is encountered. However,APDL provides a set of tools to control the programflow, such as, commands to call subroutines and tospecify branches and loops. The APDL is also used toactivate components of the Ansys graphical userinterface (GUI), such as, dialog boxes to prompt, buttons, messages and figures.

These tools allow a friendly interface to guide theuser during the whole process that involves solidgeometry modeling, mesh generation and ruptureanalysis. These tools allow the handling of the program even by users that do not have great

familiarity with this area.

Proceedings of the 8th WSEAS International Conference on APPLIED COMPUTER SCIENCE (ACS'08)

ISSN: 1790-5109 19 ISBN: 978-960-474-028-4



The PIPE program encompasses three principalmodules: input data, execute model and solve, asindicated in its initial screen showed in Fig. 1.

Fig.1. The structure of the PIPE program.

2.1 The “Input Data” ModuleIn this stage all the data necessary for the modelconstruction are introduced through a friendly

interface, composed by instructive figures and promptwindows. The geometry, finite element and materialdata are requested sequentially as the correspondentGUI tools become available.

2.1.1 Geometry Data

Initially, it is required the pipe geometric parameters:outside diameter, wall thickness and model length.

Afterwards, the user is asked about the number ofdefects to be considered in the model, taking intoaccount that there is no limitation of this number. For

none or zero defect the program switches to the nextstage and conducts the modeling of an undamaged pipe. For a number of defects higher than zero, the program generates rectangular corrosion defectslocated on the external pipe surface. For each defect, prompt boxes requesting information about its parameters and position are displayed, allowing thegeneration of multiple different rectangular defects inarbitrary position. Fig. 2 shows the screen for inputthe parameters of each defect: circumferential andlongitudinal lengths, depth and top and frontal filletradius. Fig. 3 shows the screen for input the positionof each defect: circumferential and longitudinalcylindrical coordinates of the center of the defect.

Fig.2. Screen of PIPE program for input the parameters of each defect.

Fig.3. Screen of PIPE program for input the positionof each defect.

2.1.2 Finite Element Data

In this stage the mesh density is established.Undamaged pipes are modeled with uniform meshes.Corroded pipes can be designed with a minimum of 2and a maximum of 5 regions with different meshrefinements, according to user’s choice. Each zoneincludes a region for the mesh refinement along the pipe wall thickness and a region for the meshtransition along the pipe surface, as indicated inFig.4. The distance between two adjacent pipe


ISSN: 1790-5109 20 ISBN: 978-960-474-028-4



surface transitions is calculated approximately equalto half of the maximum defect dimension, but at leastcomposed by 5 elements, due to constructiveconvenience. To automatically perform the surfacemesh refinement, each edge of this transition regioncontains a number of elements multiple of 4. Themesh refinement in the pipe surface doubles theelement side length. To perform the pipe wallthickness transition, the number of elements isattempted to be reduced to half, but limited ingenerating an even number of elements. A minimumof 2 elements along the pipe wall thickness is alsorequired.

Fig.4. Detail of the FE mesh refinement.

The PIPE program suggests a default value for theelement side length in the pipe surface to be adoptedin the first refinement level. To establish this value,the pipe wall thickness is meshed with a particularnumber of elements, multiple of 4, so that the elementside length does not exceed 3 mm. The actualelement side length along the pipe wall thickness is

then suggested as a default value for the element sidelength in the pipe surface, generating solid elementswith an aspect ratio around one. The user can modifythis value and a new mesh along the pipe wallthickness is defined.

2.1.3 Material Data

The model generated by the PIPE program allows thedefinition of only one type of material. Clicking thecorrespondent button the user is oriented to input thematerial properties using the GUI or by an input file. No default material type is available.

2.2 The “Execute Model” ModuleIn this stage, solid and mesh generations are performed simultaneously. The model is meshed withthe element Solid 45 included in the Ansys elementslibrary [4].

At the proper time, buttons indicating the possibilities to start up, to return and to continue themodeling become available to control the flow of the program.

Initially, an outline of the corroded pipe model isshowed to the user, as indicated in Fig.5. In thisstage, collapsed defects or defects out of the domainare detected and the user is oriented to start up thePIPE program clearing all data and initiating a newdata input. For an acceptable defect patchconfiguration, the user is requested to proceed the pipe modeling.

Fig.5. Screen of PIPE program showing the positionof the defects in a shell domain.

The PIPE program allows to model four layouts of

the domain, top left quarter, top half, left half andfull, simulating, respectively, the double,circumferential, longitudinal and no symmetries, asshown in Fig.6. Pressing any button the uservisualizes the correspondent domain including all thedefects within it. An option to return to the mainscreen becomes available, providing to the user theopportunity to visualize the other layouts until hisfinal decision to continue the pipe modeling.

Afterwards, a region involving all the defectsenclosed in the selected layout is delineated. Thisregion, called defect patch region, is then meshedadopting the Ansys resources of mapped or freemeshes, leading to regular or irregular meshes,


ISSN: 1790-5109 21 ISBN: 978-960-474-028-4



respectively. Regular meshes contain only 3-D solidelements with 8 nodes, as shown in Fig 7. Irregularmeshes can contain 3-D solid elements with 6 or 8nodes, as shown in Fig.8. The user can visualize bothoptions to make his decision.

Proceeding the pipe modeling, the programautomatically generates and meshes all the predefinedtransition regions and the remaining domain. Default boundary conditions are suggested, according to thelayout of the domain. The user also has the option toinput other boundary conditions.

Fig.6. Screen of PIPE program showing options ofthe domain representation.

Fig.7. FE model in the defect patch region withregular mesh.

Fig.8. FE model in the defect patch region withirregular mesh.

2.3 The “Solve” Module

If desired, the user can initiate an automatic non-linear analysis to obtain the pipe burst pressure. Thisanalysis takes into account the material non-linearityand the large strains and displacements effects.

The pipe is subjected, simultaneously, to aninternal pressure and to a longitudinal force due to the

closed ends condition. The loads are appliedincrementally until the failure criterion is reached.The non-linear analysis is performed in a controlledstep-by-step procedure using the restart resource ofthe Ansys software.

The program calculates the initial pressureincrement as being approximately equal to a quarterof the pressure value that makes the pipe reach atension level close to the material yield stress. Ateach step of the analysis the load increment is appliedin four sub-steps. The load increment is reducedwhen the maximum plastic strain exceeds the

predefined value of 0.005 or when the convergence isnot achieved within 50 iterations. The adopted failurecriterion establishes that failure is reached when thevon Mises stress along a section in the radialdirection exceeds the true ultimate tensile stress,taking into account all the points situated across thethickness. The incremental analysis stops when thisfailure criterion is attained. The analysis is alsointerrupted when the pressure increment becomeslower than 0.01 MPa. In this situation the failure pressure is considered as the maximum applied pressure load.

The Ansys will automatically quit at the end of theanalysis. The Pipe program also provides an output


ISSN: 1790-5109 22 ISBN: 978-960-474-028-4



ASCII file that records for each sub-step the pressureincrement, the number of iterations, the maximumvon Mises stress and the maximum plastic strain.

3 ValidationAlthough the finite element analysis is extremelyused to investigate the failure behavior of pipescontaining colonies of corrosion defects, its accuracyis highly dependent upon the features used in thegeneration of the model. In this section we present avalidation test for the model automatically generated by the Pipe program, demonstrating its capacity insimulating an experimental pipe burst test.

Our goal is to reproduce the results obtained in thelaboratory test of the specimen called IDTS 11reported in [5]. The tubular specimen was cut from a

longitudinal welded tube made of API 5L X80 steelwith an actual outside diameter of 459.4 mm and anactual wall thickness of 8.0 mm.

The true stress-strain curve was calculated usingthe Ramberg-Osgood equation considering the yieldstress equal to 595 MPa and the ultimate strengthequal to 738 MPa, values obtained in tensile andimpact tests registered in [5]. The true stress-straincurve is defined up to 7.11% strain. Above this valuethe stress remains constant.

The specimen IDTS 11 contains 5 defectsmachined using spark erosion. They are external

rectangular defects with smooth edges made with asmall radius of 3.2 mm. Each defect has actual width,length and depth of 32.11 mm, 40.06 mm and3.81 mm, respectively. In the photograph of thedefects of the specimen IDTS 11, shown in Fig. 9, weindicated the actual length, width and spacing of thedefects.

Fig.9. Photo of the specimen IDTS 11 showing thelength, width and spacing of the defects.

The solid FE model was constructed automatically by the PIPE program. Due to the position of thedefects, the left half of the domain was modeled andthe appropriate default boundary conditions wereapplied to simulate the longitudinal symmetry. Themodel was designed with 4 mesh refinement levels.In the more refined region the element side lengthadopted in the pipe surface was 1.5 mm, which wasincreased to 3.0 mm, 6.0 mm and 12 mm in theneighboring regions. In the defect patch region, 4elements were used through the pipe wall thickness.In the remaining domain 2 elements were usedthrough the pipe wall, since it is the minimumnumber required by the program. A detail of the solidFE model can be visualized in Fig. 10.

Fig.10. Detail of the solid FE model of IDTS 11.

In Fig 11 we present the von Mises stressdistribution, in the region of the defects, obtained bythe FE non-linear analysis at the burst pressure levelof 20.59 MPa. This pressure level was achievedaccording to the adopted failure criterion, whichestablishes that failure occurs when the von Misesstresses of all the points situated along a cross section

in the radial direction reach the true ultimate tensilestress.

In the numerical analysis of the specimenIDTS 11, the failure process develops from the pipetop surface to the pipe bottom surface. The placewhere the failure occurs is indicated in Fig 11. It can be observed that, the predicted failure configurationis in agreement with that obtained in the laboratorytest shown in Fig. 12.

In Table 1, the failure pressure measured in thelaboratory test [5] is compared to those predicted bythe assessment methods ASME B31G [6] and DNVRP-F101 [7] and with the result obtained in the AnsysFE analysis performed with the PIPE program. In the


ISSN: 1790-5109 23 ISBN: 978-960-474-028-4



same table, the errors of the failure pressure predictions are presented. According to this table allthe predicted failure pressures are conservative,showing values smaller than the actual failure pressure. The most accurate result is the one provided by the FE analysis, showing its reliability.

Fig.11. von Mises stresses distribution in the defectsregion at an internal pressure load of 20.59 MPa.

Fig.12. Photo of the defects of the specimen IDTS 11after failure.

Table 1. Actual and predicted failure pressures andcorrespondent errors

MethodFailure Pressure

(MPa)Error (%)

Burst test 21.260 0FEM 20.590 -3.15DNV 18.714 -11.97

B31G 17.756 -16.48Error(%)=((predicted-experimental)/experimental)*100

4 ConclusionIn this paper we presented the principal tools of thePIPE program developed to generate solid finiteelement models and to manage FE analyses ofundamaged and corroded pipelines, automatically.

The program provides a friendly graphical userinterface for the Ansys software allowing a guidedand quick modeling of pipes containing multipledifferent defects in arbitrary position. A validationtest was presented showing that the program leads toan appropriate model generation and to a reliablenumerical simulation.

References:

[1] R.C.C. Silva, J.N.C. Guerreiro and A.F.D. Loula,A Study of Pipe Interacting Corrosion Defects

Using the FEM and Neural Networks, Advancesin Engineering Software, No.38, 2007, pp. 868-875.

[2] E.Q. Andrade, A.C.Benjamin, P.R.S. Machado Jr,L.C.Pereira, B.P. Jacob, E.G. Carneiro, J.N.C.Guerreiro, R.C.C. Silva and D.B. Noronha Jr,Finite Element Modeling of the Failure Behaviorof Pipelines Containing Interacting CorrosionDefects, Proc. 25

th International Conference on

Offshore Mechanics and Arctic Engineering,OMAE 2006.

[3] A.C.Benjamin, D.J.S. Cunha, R.C.C. Silva,

J.N.C. Guerreiro, G.C. Campello and F.E. Roveri,Stress Concentration Factors for a Drilling RiserContaining Corrosion Pits, Proc. 26

th

International Conference on Offshore Mechanics

and Arctic Engineering, OMAE 2007.[4] Ansys, Ansys Release 9.0 Documentation, 2004.[5] A.C.Benjamin, R.D.Vieira, J.L.F. Freire and

E.Q.Andrade, Burst Testes on Pipeline ContainingClosely Spaced Corrosion Defects, Proc. 25

th

International Conference on Offshore Mechanics

and Arctic Engineering, OMAE 2006.[6]Anon, ASME-B31G – Manual for Determining

the Remaining Strength of Corroded Pipelines –A Supplement to ANSI/ASME B31 Code forPressure Piping, American Society of Mechanical

Engineers, New York,1991.[7]Anon, DNV Recommended Practice - DNV – RP-

F101 – Corroded Pipelines, Det Norske Veritas, Norway,1999.


ISSN: 1790-5109 24 ISBN: 978-960-474-028-4

corrosion assessment

Documents