1555_dowhopoluk

Upload: idiazg

Post on 14-Apr-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/29/2019 1555_Dowhopoluk

    1/73

    04.09.13

    Advanced Topics in Buried Analysis

    Michael Dowhopoluk, M.Eng., P.Eng.

    PVP Engineering Ltd.

  • 7/29/2019 1555_Dowhopoluk

    2/73

    04.09.13

    Advanced Topics in Buried Analysis

    2

  • 7/29/2019 1555_Dowhopoluk

    3/73

    04.09.13

    Overview Todays Topics

    Overview of Buried Pipeline Codes and Design

    Pipeline Stress Analysis

    Different modeling techniques

    Nonlinear Stress Analysis / Limit States Design

    Nonlinear analysis with linear tools

    Total strain theory of plasticity

    Validation and Verification

    4

  • 7/29/2019 1555_Dowhopoluk

    4/73

    OVERVIEW OF BURIED PIPELINECODES AND DESIGN

  • 7/29/2019 1555_Dowhopoluk

    5/73

    04.09.13

    Overview

    Applicable Codes

    B31.4 Liquid Petroluem First appeared in 1959

    B31.8 Gas Transmission First appeared in 1958

    CSA Z183 Oil Transmission 1967

    CSA Z184 Gas Transmission 1968

    CSA Z662 Replaced Z183 / Z184 in 1994

    6

  • 7/29/2019 1555_Dowhopoluk

    6/73

    04.09.13

    Buried Pipeline Design

    Code Requirements

    Hoop Stress

    Expansion Stress

    Combined Stress

    Additional Loadings

    Contrast to B31.3

    No combined limit why not?

    Limit on gross deformation due to yielding

    Only a concern when there are large axial compressive forces

    7

  • 7/29/2019 1555_Dowhopoluk

    7/73

    04.09.13

    B31.4 Liquid Petroleum Pipelines

    Basic Allowable Stress

    = 0.72 ; , 1.0

    Hoop Stress

    =

    Longitudinal Stress

    =

    + 0.75 = 0.525

    Thermal Expansion Stress Range

    Restrained =

    = Max , 0.9

    8

  • 7/29/2019 1555_Dowhopoluk

    8/73

    04.09.13

    B31.4 Liquid Petroleum Pipelines

    Thermal Expansion Stress Range

    Unrestrained

    = + 4 0.7

    9

  • 7/29/2019 1555_Dowhopoluk

    9/73

    04.09.13

    B31.8 Gas Transmission Pipelines

    Basic Allowable Stress

    = ; , 0.72 0.8

    Hoop Stress

    =

    Thermal Expansion Stress Range

    = + 4 0.72

    10

  • 7/29/2019 1555_Dowhopoluk

    10/73

    04.09.13

    CSA Z662 Oil and Gas Pipelines

    Basic Allowable Stress

    = ; ,0.8

    Hoop Stress

    =

    Longitudinal Stress

    =

    +

    Combined Hoop and Longitudinal

    Restrained =

    0.9

    11

  • 7/29/2019 1555_Dowhopoluk

    11/73

    04.09.13

    CSA Z662 Oil and Gas Pipelines

    Combined Stress for Restrained Spans

    +

    Thermal Expansion Stress Range

    = + 4 0.72

    12

  • 7/29/2019 1555_Dowhopoluk

    12/73

    PIPELINE STRESS ANALYSIS

  • 7/29/2019 1555_Dowhopoluk

    13/73

    04.09.13

    Analysis

    Buried pipeline can be classified into fully restrained, partially restrained, and

    unrestrained sections

    Fully restrained => plane strain problem, no bending

    Longitudinal stresses, but no strains

    Can be analysed by hand calculations

    Partially restrained/Unrestrained => 3D stress state, often approximated as

    plane stress

    FEA required to capture nonlinear soil response ie. CAESAR

    Axial elongation due to pressure and thermal loads

    Lets try a sample model

    14

  • 7/29/2019 1555_Dowhopoluk

    14/73

    04.09.13

    Sample Problem

    Single 15 Bend, Virtual Anchor on each side

    CSA Z662 design

    NPS 12, 7.9mm (.311 in), Gr. 414 (X60)

    Design Pressure 9.93 MPa (1440 psi)

    Design Temperature - 60C (140F)

    Install Temperature - -25C (-13F)

    Depth of Burial 1500mm (5 ft)

    15

  • 7/29/2019 1555_Dowhopoluk

    15/73

    04.09.13

    Basic Code Calculations

    16

  • 7/29/2019 1555_Dowhopoluk

    16/73

    04.09.13

    CAESAR Model

    Five elements

    15D Induction Bend

    17

  • 7/29/2019 1555_Dowhopoluk

    17/73

    04.09.13

    Soil Model

    ALA method preferred

    Detailed soils information usually not available

    Best Practice

    Parametric Study

    Minimum of four soil models

    Cohesive and non-cohesive

    High/low compaction

    Extra caution required for:

    Frozen Backfill

    High Water Table / Muskeg

    Slopes

    18

  • 7/29/2019 1555_Dowhopoluk

    18/73

    04.09.13

    Results

    19

  • 7/29/2019 1555_Dowhopoluk

    19/73

    04.09.13

    Discussion

    All Done!

    Lets call ourselves pipeline experts and all go home.

    But,.perhaps a little Validation and Verification before we go.

    Hoop Stress

    = /2 = 323.9 * 9.93 / (2*7.9) = 203.6 MPa vs. 193.6 MPa

    Restrained Longitudinal Stress

    = 1 = 1 .3 203.6 + 2.075 1.2 5 60 25 = 353.7 MPa vs. 327.5 MPa

    Allowable Stress

    Fully restrained should be 0.9*Sy = 372.6 MPa vs 414 MPa

    20

  • 7/29/2019 1555_Dowhopoluk

    20/73

    04.09.13

    Hoop Stress

    What is being reported in the Sustained case?

    Not hoop stress, but the longitudinal stress due to pressure

    ie.1

    = 0.5 * 193.6

    = 96.8 MPa 97.3 MPa

    Revise model inputs

    Hoop Stress Calc based on OD

    Mill tolerance not required for CSA

    Change Poissons ratio to 0.3

    21

  • 7/29/2019 1555_Dowhopoluk

    21/73

    04.09.13

    Hoop Stress

    22

    One Down!

  • 7/29/2019 1555_Dowhopoluk

    22/73

    04.09.13

    Longitudinal Stress

    Revise some more inputs

    Replace temperature with thermal strain

    Revise elastic modulus

    Re-run the analysis

    24

  • 7/29/2019 1555_Dowhopoluk

    23/73

    04.09.13

    Longitudinal Stress

    25

    Two Down!

  • 7/29/2019 1555_Dowhopoluk

    24/73

    04.09.13

    One more thing.Pressure Thrust

    27

    +20% !

    Why did the restrained

    stress go up?

  • 7/29/2019 1555_Dowhopoluk

    25/73

    04.09.13

    Pressure Loading

    By definition, pressure acts in all directions

    For a closed cylinder, there are stresses and strains in all 3 directions

    When the end cap forces are present, the general treatment is,

    =

    /4

    = /2

    = ,, generally neglected

    =1

    +

    1/

    (

    )

    = 1 + 1/

    (

    )

    =1

    +

    3

    (

    )

    28

  • 7/29/2019 1555_Dowhopoluk

    26/73

    04.09.13

    Pressure Loading

    What happens when end cap forces are not present?

    Eg. Expansion joint

    = 0 = /2

    = 0 =

    1

    +

    (

    )

    =1

    +

    1

    (

    )

    =1

    +

    (

    )

    29

  • 7/29/2019 1555_Dowhopoluk

    27/73

    04.09.13

    Pressure and Thermal Loads

    Thermal expansion causes a stress free strain, unless it is restricted.

    = ; = 0

    = ; = 0

    Thus, we can solve for the restrained stress,

    + = 0 =1

    + +

    =

    For the unrestrained case

    = 0

    30

  • 7/29/2019 1555_Dowhopoluk

    28/73

    04.09.13

    Pressure Loading

    How are the end cap loads treated in CAESAR?

    As a uniform tensile (+ve) strain applied over the entire system

    =1

    +

    1/

    (

    )

    For restrained sections, = 0, this results in a compressive stress

    In actuality, it should be a tensile force and a ve strain

    This method is computationally attractive, but has some drawbacks

    Point of application of the force is lost; this leads to an incorrect distribution of

    forces, stresses, and strains in the pipeline model

    Compressive stresses in the restrained portions of the pipeline are over-reported.

    Lets look at this in more detail

    31

  • 7/29/2019 1555_Dowhopoluk

    29/73

    04.09.13

    Pressure Loading

    Returning to the sample problem,

    =

    =

    ..3

    . 9.93

    33.8

    .9= .02% ; = 40.7 ;

    + = 353.7 + 40.7 = 394.4 . 391.3

    Now we know what CAESAR is calculating

    How can we go about improving this model?

    32

  • 7/29/2019 1555_Dowhopoluk

    30/73

    04.09.13

    Pressure Loading PVP Method

    Step #1 Add thrust forces at the correct locations

    Thrust force = ()

    Use the near and far points on the bend

    Bend mid-point is not advisable

    Step #2 Convert to an equivalent strain; subtract from thermal strain

    Set Poissons ratio to a very small value (0.001)

    Step #3 Post-process the bend manually

    Introduction of thrust forces generates compressive forces in the bend that are notreal

    Pressure thrust acts on the projected area of the pipe tangents, on the backside of

    the bend; there is no net section compression

    33

  • 7/29/2019 1555_Dowhopoluk

    31/73

    04.09.13

    Step #1 Add Thrust Forces

    Thrust Force = = 9.93 (

    33.9

    7.9)= 740 !

    34

  • 7/29/2019 1555_Dowhopoluk

    32/73

    04.09.13

    Step 2 Poisson Effect

    Calculate the compressive strain/ = -.3*203.5/2.07e5 = .000295

    Subtract from thermal strain (.00102 - .000295 = .000725)

    Note that the Poisson strain can be converted to an

    equivalent temperature by dividing by the co-efficient of

    thermal expansion (ie. .000295/1.2e-5 = 24.5C)

    Now, run the analysis!

    Note the load cases Ive

    set up.

    35

  • 7/29/2019 1555_Dowhopoluk

    33/73

    04.09.13

    Step 3 Post-process

    Take the local axial forces for the bend in the OPE (W+T1+P1+F) case

    Subtract the local axial force over the bend from the OPE (F) case

    Manually calculate the axial and bending stresses

    Excel time!

    36

  • 7/29/2019 1555_Dowhopoluk

    34/73

    04.09.13

    Post Processing - Excel

    37

  • 7/29/2019 1555_Dowhopoluk

    35/73

    04.09.13

    Results Discussion

    What do you end up with?

    Maximum OPE stress of 571 MPa vs 704 MPa (23% reduction)

    Displacements 36mm vs 44mm (23% reduction)

    Restrained Axial Force 1176 kN vs 1951 kN

    Restrained Stress 353 MPa vs 391 MPa

    Is it worth the effort?

    For the right problem, yes.

    38

  • 7/29/2019 1555_Dowhopoluk

    36/73

    04.09.13

    Results Discussion

    Axial force in the line

    39

  • 7/29/2019 1555_Dowhopoluk

    37/73

    04.09.13

    A little V&V

    Those numbers look good, but.

    weve always used the CAESAR defaults with/without Bourdon

    the CAESAR method is the industry standard

    weve never had any issues

    We thought the same, so..

    PVP developed an ANSYS APDL Pipeline design suite.

    Linear/nonlinear steel plasticity models

    Large deformation theory

    New generation pipe elements including shell deformation modes

    We benchmarked the same problem using ANSYS

    40

  • 7/29/2019 1555_Dowhopoluk

    38/73

    04.09.13

    ANSYS Results

    41

    Results Agreement -

    2% on Stress

    10% on Displacement

  • 7/29/2019 1555_Dowhopoluk

    39/73

    NONLINEAR STRESS ANALYSIS

  • 7/29/2019 1555_Dowhopoluk

    40/73

    04.09.13

    Nonlinear Pipeline Design

    What is Nonlinear Analysis?

    Metal Plasticity

    Geometric (ie. bend ovalization)

    Path dependent loading

    Why bother for a Pipeline?

    Reduction in capital cost

    Better defined safety margin

    Failure assessment

    43

  • 7/29/2019 1555_Dowhopoluk

    41/73

    04.09.13

    Cost Reduction

    Comparison of Required Wall Thickness

    NPS 12, Gr. 483, 9.93 MPa

    44

  • 7/29/2019 1555_Dowhopoluk

    42/73

    04.09.13

    Cost Reduction

    Substitution of Lower Grade

    NPS 12, 7.9mm, 9.93 MPa

    45

  • 7/29/2019 1555_Dowhopoluk

    43/73

    04.09.13

    Safety Margin

    Higher stress = increased safety margin ?

    Pipelines fail for many reasons.

    (corrosion, external damage, denting, cracking)

    Does a single allowable stress appropriately address these failure modes?

    Limit states design

    Reliability based design method based on factored loads and factored resistances

    Based on identification of failure modes

    Assignment of appropriate safety factors for Ultimate, Serviceability

    Conclusion

    Better defined safety margin

    Remove waste in the design from by using a consistent safety margins

    46

  • 7/29/2019 1555_Dowhopoluk

    44/73

    04.09.13

    Plasticity

    Theory of strain past yield

    Two main theories Incremental and Total

    Incremental

    Increments of plastic strain related to plastic

    stress by the tangent modulus

    Overall response is the integral over the

    load path

    Total (Secant)

    Total strain related to total stress by thesecant modulus

    48

  • 7/29/2019 1555_Dowhopoluk

    45/73

    04.09.13

    Incremental Theory

    More widely known and implemented (ANSYS, ABAQUS, etc)

    Capable as a general theory

    More computationally intensive

    = (, )

    Much more complex than we have

    time for.

    49

  • 7/29/2019 1555_Dowhopoluk

    46/73

    04.09.13

    Total strain theory

    Total strains related to total stresses by secant modulus

    For a certain class of problems its much easier to apply

    Restrictions

    Stress history effects cannot be accounted for

    Proportional loading only

    Isotropic hardening only

    Lets investigate this further

    50

  • 7/29/2019 1555_Dowhopoluk

    47/73

    04.09.13

    Nonlinear Material Models

    Many well known models to choose from

    Bilinear

    = ;

    = 1 + 1 ; >

    Ramberg-Osgood

    =

    + (

    ) ;

    = 10. . 30

    ASME VIII

    Multi-parameter

    51

  • 7/29/2019 1555_Dowhopoluk

    48/73

    04.09.13

    Yield Criteria

    Two main theories

    Maximum Shear Stress

    ( ) =

    Equivalent Stress

    1

    1 + 3

    + 3 1 =

    52

  • 7/29/2019 1555_Dowhopoluk

    49/73

    04.09.13

    Yield Surfaces

    53

  • 7/29/2019 1555_Dowhopoluk

    50/73

    04.09.13

    Hardening Laws

    What happens when you load a material past yield?

    Elastic-Plastic

    No change in stress

    Strain is undefined

    Isotropic

    Uniform expansion of yield surface

    Same center point

    Kinematic

    A shift in the yield surface

    Same size

    54

  • 7/29/2019 1555_Dowhopoluk

    51/73

    04.09.13

    Plastic Incompressibility

    Elastic deformation is compressible

    Obvious since Poissons ratio is < , usually about 0.3

    Plastic deformation is incompressible

    Less obvious, had been assumed; proven experimentally by Bridgman

    Poisson ratio equal to 0.5

    What about the range in the middle?

    There is a transition, =

    It can be proven that the relation is as follows

    =1

    [

    1

    ]

    55

  • 7/29/2019 1555_Dowhopoluk

    52/73

    04.09.13

    Load Path Dependence

    Ex. - Thin Cylinder with Tension and Torsion

    58

    Initial Yield Curve

    Yield Curve after

    Hardening

  • 7/29/2019 1555_Dowhopoluk

    53/73

    04.09.13

    Load Path Dependence

    Ex. Thin Cylinderwith Tensile Loading

    Load along OABAO

    Yield at point A, strain harden to point B, remove load ( = 0)

    = ;

    =

    = ;

    = 0

    59

    Initial Yield Curve

    Yield Curve after

    Hardening

  • 7/29/2019 1555_Dowhopoluk

    54/73

    04.09.13

    Load Path Dependence

    Ex. Thin Cylinderwith Torsional Loading

    Load along OCDCO

    Yield at point C, strain harden to point D, remove load ( = 0)

    = 0 ;

    = ;

    =

    = 0

    60

    Initial Yield Curve

    Yield Curve after

    Hardening

  • 7/29/2019 1555_Dowhopoluk

    55/73

    04.09.13

    Load Path Dependence

    Ex. Arbit rary Loading

    Two different load paths - OABAEF and OCDF.

    Same final stress state. Different strains.

    What are they?

    61

    Initial Yield Curve

    Yield Curve after

    Hardening

  • 7/29/2019 1555_Dowhopoluk

    56/73

    04.09.13

    Load Path Dependence

    Ex. Arbit rary Loading

    OABAEF -

    = ;

    =

    =

    OCDF -

    =

    They are unchanged. All unloading/reloading is elastic!

    62

    Initial Yield Curve

    Yield Curve after

    Hardening

  • 7/29/2019 1555_Dowhopoluk

    57/73

    04.09.13

    Load Path Dependence

    Ex. Deformation Plasticity

    Recall that it only can handle proportional loading.

    Path OF is proportional

    63

    Initial Yield Curve

    Yield Curve after

    Hardening

  • 7/29/2019 1555_Dowhopoluk

    58/73

    04.09.13

    CAESAR Nonlinear Analysis

    Step #1 Linear solution

    Step #2 Extract Element Stress

    Step #3 Calculate effective modulus (E) and effective Poisson ratio (v)

    Step #4 Repeat Steps #2 and #3 by element

    Step #5 Solve

    Step #6 Compare change in stress to convergence criteria. If converged,

    then stop, else Step #2.

    64

  • 7/29/2019 1555_Dowhopoluk

    59/73

    04.09.13

    CAESAR Nonlinear Analysis

    Step #1 Start with Solution from PVP Method

    Step #2 Extract Element Stress

    Select range of elements with stress above yield

    Step #3 Calculate E and v

    Use material model to calculate E and v

    Step #4 Repeat by Element

    Step #5 - Solve

    65

  • 7/29/2019 1555_Dowhopoluk

    60/73

    04.09.13

    CAESAR Nonlinear Analysis

    Compare change in stress to convergence criteria (eg. 5%)

    Repeat.repeat.repeat

    66

    Iteration #2Iteration #1 Iterations #3,4,5,6

  • 7/29/2019 1555_Dowhopoluk

    61/73

    04.09.13

    CAESAR Nonlinear Analysis

    Max + , Bend Midpoint

    Iteration #1 571 MPa

    Iteration #2 234 MPa

    Iteration #3 494 MPa

    Iteration #4 461 MPa

    Iteration #5 485 MPa

    Iteration #6 477 MPa

    We have convergence to 2%, and have achieved a 20% stress reduction

    Based on the material model, we have 0.2% plastic strain.

    67

  • 7/29/2019 1555_Dowhopoluk

    62/73

    04.09.13

    Results Discussion

    Is it worth the effort?

    Quite laborious and time consuming

    Lots of data manipulation, error prone.

    But, for the right problem,

    You get a much better understanding of the actual stress in the line.

    You are not forced to make design changes due to the limitations of

    your analysis method.

    That all looks very nice, but have you validated the results?

    Of course we did.

    68

    ANSYS N li

  • 7/29/2019 1555_Dowhopoluk

    63/73

    04.09.13

    ANSYS Nonlinear

    Kinematic Hardening, Small Def. Theory, Pipe289 Elements

    69

    Agreement

    to 2.5%!

    ANSYS N li

  • 7/29/2019 1555_Dowhopoluk

    64/73

    04.09.13

    ANSYS - Nonlinear

    Kinematic Hardening, Small Def. Theory, Pipe289 Elements

    70

    Agreement

    to 5%.

    A W d f C ti

  • 7/29/2019 1555_Dowhopoluk

    65/73

    04.09.13

    A Word of Caution

    There are limits to the capabilities of the beam theory elements

    Key problem is the assumption that cross-sections remain plane and

    undistorted after deformation

    Not capable of capturing ovalization, wrinkling, or collapse

    Ovalization has a pronounced effect at higher strain levels

    Be wary of problems with high D/t

    Initial imperfections play a large role in wrinkle and buckle initiation

    We do not consider the method to be sufficiently robust to do a full strain-

    based design.

    So what is it good for?

    71

    CAESAR N li

  • 7/29/2019 1555_Dowhopoluk

    66/73

    04.09.13

    CAESAR Nonlinear

    Stress based design!

    There is a lot to be gained

    between 70% SMYS and

    90% SMYS

    Codes dont say you cant vary

    E and v.

    About .1% more strain

    allowed

    =

    1

    E

    72

  • 7/29/2019 1555_Dowhopoluk

    67/73

    RISER DESIGN

    D i Phil h

  • 7/29/2019 1555_Dowhopoluk

    68/73

    04.09.13

    Design Philosophy

    Why is there an issue?

    Numerous failures

    Reasons difficult to model

    Backfill conditions

    Compaction

    Local stresses

    Restrained or Unrestrained?

    Design Robustness

    74

    Backfill

  • 7/29/2019 1555_Dowhopoluk

    69/73

    04.09.13

    Backfill

    Can be somewhat different than what you model.

    75

    Piled Risers

  • 7/29/2019 1555_Dowhopoluk

    70/73

    04.09.13

    Piled Risers

    The outcome of multiple lawsuits.

    About the worst stress design you

    could come up with.

    Lets discuss

    76

    Piled Risers

  • 7/29/2019 1555_Dowhopoluk

    71/73

    04.09.13

    Piled Risers

    And a common outcome.

    77

    Restrained Risers

  • 7/29/2019 1555_Dowhopoluk

    72/73

    04.09.13

    Restrained Risers

    Restrained = Anchor Block

    Usually concrete due to loadrequirements.

    Points to Consider

    Familiarity with the design and

    installation of an anchor block

    Design of the attachment to the pipe Protection of pipe coating

    Availability of concrete

    Load calculation

    Cost

    Recommendation

    Least preferred option

    78

    Unrestrained Risers

  • 7/29/2019 1555_Dowhopoluk

    73/73

    Unrestrained Risers

    Unrestrained is a matter of degree

    Shallow angle riser Buried Offset

    Points to Consider

    Extra space requirements

    May required controlled backfill

    compaction May require imported backfill

    Generally more economical

    Easiest to implement early in

    design process

    Recommendation This is my preferred riser solution