a finite element approach to solve contact problems in geotechnical engineering

Upload: anh-do

Post on 06-Apr-2018

217 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    1/26

    INTERNATIONAL JOURNAL FOR NUMERICAL AND ANALYTICAL METHODS IN GEOMECHANICSInt. J. Numer. Anal. Meth. Geomech., 2005; 29:525550Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/nag.424

    A finite element approach to solve contact problems in

    geotechnical engineering

    Jianqiang Maon,y

    Department of Geotechnical Engineering, Southwest Jiaotong University, Chengdu, Peoples Republic of China

    SUMMARY

    This paper presents a finite element approach to solve geotechnical problems with interfaces. Thebehaviours of interfaces obey the MohrCoulomb law. The FEM formulae are constructed by means ofthe principle of virtual displacement with contact boundary. To meet displacement compatibilityconditions on contact boundary, independent degrees of freedom are taken as unknowns in FEM

    equations, instead of conventional nodal displacements. Examples on pressure distribution beneath a rigidstrip footing, lateral earth pressure on retaining walls, behaviours of axially loaded bored piles, a shield-driven metro tunnel, and interaction of a sliding slope with the tunnels going through it are solved withthis method. The results show good agreement with analytical solutions or with in situ test results.Copyright# 2005 John Wiley & Sons, Ltd.

    KEY WORDS: interface; contact problem; principle of virtual displacement; finite element; geotechnical

    1. INTRODUCTION

    The interfaces, such as interface between the structure and geotechnical media, the joints in rock

    mass, failure surfaces in soil mass, etc., effect the mechanical behaviours of the structures and

    the surrounding geotechnical media significantly. Various numerical interface models have been

    developed by Goodman et al. [1], Zienkiewicz et al. [2], Clough and Duncan [3], Ghaboussi et al.

    [4] and Desai et al. [5]. These models or their derivative forms have been widely applied in

    geotechnical engineering, as they can be readily incorporated into an FEM process. However, a

    common and distinct disadvantage of these models is that additional constitutive equations and

    mechanical parameters have to be employed for the interfaces.

    As we know, in spite of the variety of interfaces, the following characters are essential for

    an interface model, as shown in Figure 1: (1) if the tangential force at a point on interface

    reaches the limiting value of resistance, relative slip will occur, else will keep in sticking status

    (2) whether relative slip occurs or not, the contact bodies cannot penetrate each other in

    any way.

    Received 21 January 2004Revised 10 November 2004Copyright# 2005 John Wiley & Sons, Ltd.

    yE-mail: [email protected]

    nCorrespondence to: J. Mao, Department of Geotechnical Engineering, Southwest Jiaotong University, Chengdu,Peoples Republic of China.

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    2/26

    In fact, almost all the problems containing interfaces can be uniformly described as a contact

    problem through the following mathematical form: assuming the two contact bodies are O and

    O0; respectively, and on the contact boundary (i.e. the interface) Gc:

    (1) n is the normal direction 2 O and s is the slipping direction. Accordingly, es and es0 are

    unit vectors corresponding to s and s0 (opposite to s: Also we note that for two-

    dimensional problems, the possible slipping direction is uniquely along the tangential

    direction, while for three-dimensional problems, the slip will occur in the direction of

    maximum tangential force, which will depend on the contact forces at the point.

    (2) The contact forces on O0 are pn and ps; and on O are p0n and p0s:(3) The displacements in incremental form (for contact problems are nonlinear) are du0n; du

    0s

    and dun; dus: Moreover, d0n is the initial gap between contact bodies (for most of

    geotechnical problems, d0n 0).

    (4) The slip on the interface is govern by the MohrCoulomb law

    jpsj cB mjpnj 1

    where cB and m are cohesion and coefficients of friction of the interface, respectively.

    When slip occurs, the cohesion at the slipping point will be lost, i.e. cB 0: The

    conditions to be satisfied on interface are summarized in Table I.

    There are many contact problems in mechanical and civil engineering, hence numerous efforts

    has been made in last few decades to develop corresponding numerical algorithms. A number oftechniques such as iteration method [68], mathematics programming [912], penalty method

    [13], Lagrange multiplier [1416], have been developed and used in mechanical engineering

    problems. However, these methods are rarely applied in geotechnical engineering [17, 18] due to

    their complex techniques in computation. A practical approach applicable to geotechnical

    problems is presented in this paper.

    np

    np

    np

    np

    sps

    p

    npn p

    relative slip along tangential direction

    contact status when slip does not occur

    ns

    cannot penetrate each other in normal direction

    contact status when slip occurs

    a pair of contact points

    Figure 1. Illustration of the contact forces and displacements on interface.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO526

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    3/26

    2. FEM FORMULATION BASED ON THE PRINCIPLE OF VIRTUAL WORK

    WITH CONTACT BOUNDARY

    2.1. Principle of virtual work for contact problems

    The principle of virtual work is one of the important principles in analytical mechanics

    concerning the kinematics of particles or rigid bodies. For a deformable body, if the body is

    divided into a series of infinitesimal elements and is taken as particles in sense of kinematics,

    then the whole body will lead to an assembly of particles.

    Figure 2 illustrates three different types of infinitesimal elements: (1) in bodies O and O0; (2) on

    stress boundaries Gs and G0s; and (3) on contact boundary Gc: Gc consists of sticking part Gc1 and

    slipping part Gc2 ; which Gc Gc1 [ Gc2 : The infinitesimal elements are same as those used in

    continuous mechanics to construct the equations of equilibrium. The expressions of resultant

    forces acting on different infinitesimal elements are summarized in Table II.

    According to the principle of virtual displacement, the virtual work of whole system should be

    equal to zero. So we obtainZO

    r r %FF du dO

    ZO

    0

    r0 r %FF0 du0 dO

    ZGs

    %pp p du dG

    ZG

    0s

    %pp0 p0 du0 dG

    ZGc1

    pn p0n dun ps p

    0s dus

    p0n pn du0n p

    0s ps du

    0s dG

    ZGc2

    pn p0n dun p

    0n pn du

    0n ps mjp

    0njes0 dus

    p0s mjpnjes du0s dG 0 2

    where du; du0 are virtual displacements which should satisfy the displacement conditions on

    displacement boundary Gu; G0u and contact boundary Gc in prior, i.e.

    du 0 on Gu; du0 0 on G0u 3a

    Table I. Conditions to be satisfied on interface.

    Traction DisplacementContact status

    p p0 s n s

    Sticking or ps5cB mjpnjes dun du0n

    dus du0s

    pn p0n

    p0s5cB mjp0njes0

    Slipping and ps mjpnjes jun u0nj d

    0n dus=du

    0s

    ps p0s p

    0s mjp

    0njes0

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 527

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    4/26

    dun du0n; dus du

    0s on Gc1 3b

    dun du0n; dus=du

    0s on Gc2 3c

    Figure 2. Stresses and resultant forces on infinitesimal elements.

    Table II. Resultant forces acting on infinitesimal elements.

    Infinitesimal Direction of the Resultant of the forceselement resultant force on infinitesimal element Virtual work

    In O x1; x2; x3 r r %FF dO r r %FF du dO

    In O0 x1; x2; x3 r0 r %FF0 dO r0 r %FF0 du0 dO

    On Gs x1; x2; x3 %pp p dG %pp p du dG

    On G0s x1; x2; x3 %pp0 p0 dG %pp0 p0 du0 dG

    On Gc1 (to O) n pn p0n dG pn p

    0n dun dG

    s ps p0s dG ps p

    0s dus dG

    On Gc1 (to O0) n0 p0n pn dG p

    0n pn du

    0n dG

    s0 p0s ps dG p0s ps du

    0s dG

    On Gc2 (to O) n pn p0n dG pn p

    0n dun dG

    s ps mjp0njes0 dG ps mjp

    0njes0 dus dG

    On Gc2 (to O0) n0 p0n pn dG p

    0n pn du

    0n dG

    s0 p0s mjpnjes dG p0s mjpnjes du

    0s dG

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO528

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    5/26

    Substituting Equations (3b), (3c) and

    ZO

    r r du dO

    ZG

    p du dG

    ZO

    r : de dO

    Z

    G

    pn dun ps dus dG Z

    O

    r : de dO 4a

    ZO

    r0 r du0 dO

    ZG

    p0 du0 dG

    ZO

    r0: de0 dO

    ZG

    p0n du0n p

    0s du

    0s dG

    ZO

    r0: de0 dO 4b

    into Equation (2), finally we get

    ZO

    r : de dO Z

    O0r0 : de0 dO

    ZO

    %FF du dO Z

    O0 %FF0 du0 dO

    ZGs

    %pp du dG

    ZG

    0s

    %pp0 du0 dG

    ZGc1

    pn p0n dun ps p

    0s dus dG

    ZGc2

    pn p0n dun

    mjp0njes0 dus mjpnjes du0s dG 5

    The above equation form the basis to construct the FEM formulae in the next section.

    In the case that one of the bodies, such as O0; is stiff enough to be taken as a rigid body, the

    virtual displacement of a generic point in O0 may be expressed as

    du0 dV0 R0 dH0 A0dV0

    dH0

    ( ) A0 dU0 6

    where V0 is the translation of the rigid body, and H0 is the rotation. R0 is not a tensor and its

    components are

    R0

    ij

    X3

    k1

    eijkx0

    k

    where eijk is Ricci Symbol, x0k k 1; 2; 3 is co-ordinate of the point (the origin is designated at

    the centre of rotation).

    The equation corresponding to rigid body-deformable body contact can be obtained by

    substituting Equation (6) into (5). The expression is omitted here on.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 529

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    6/26

    2.2. Finite element equilibrium equations

    After the discretization, Equation (5) takes the following form:

    XZO

    e

    O0e

    BTr dO XZO

    e

    O0e

    NT %FF dO XZG

    e

    sG0e

    s

    NT %pp dG

    XZ

    Gec1

    NTnpn N0Tn p

    0n dG

    XZ

    Gec1

    NTs ps N0Ts p

    0s dG

    XZG

    ec2

    NTnpn N0Tn p

    0n

    XZ

    Gec2

    NTs mjpnj N0Ts mjp

    0nj dG 7

    where B is straindisplacement matrix, N (and N0 is displacement interpolation matrix, and

    similarly Ns and N0sNn and N

    0n are tangential (normal) displacement interpolation matrices.

    Direct solution of equilibrium equation (7) is a very complicated procedure, hence somesimplifications are made.

    In Equation (7), tractions pn;ps and p0n;p

    0s on contact boundary G

    ec1

    are computed with the

    stresses of different elements Oe and O0e; respectively, and are not equal. To simplify the

    computation, takingPR

    Gc1NTnpn;

    PRGc1

    NTs ps andPR

    Gc1N0Tn p

    0n;PR

    Gc1N0Ts p

    0s as nodal contact

    forces corresponding to pn; ps and p0n; p

    0s on Gc1 ; as shown in Figure 3, we can assume

    approximately XZG

    ec1

    NTnpn N0Tn p

    0n dG

    XG

    ec1

    Pni P0ni 0 on Gc1

    XZG

    ec1

    NT

    sp

    s N0T

    sp0

    s dG X

    Gec1

    Psi

    P0

    si 0 on G

    c1

    niPniP

    siP

    siP

    n

    s

    e

    e

    e

    e e

    e

    ( )ni sP

    ( )ni sPniP

    iniP

    2

    ec1

    ec

    (a) (b) (c)

    Figure 3. Nodal contact forces on contact boundary: (a) Definition ofn and s at contact node i; (b) contactforces on Gec1 ; and (c) contact forces on G

    ec2:

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO530

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    7/26

    Similarly, we obtainXZG

    ec2

    NTnpn N0Tn p

    0n dG

    XG

    ec2

    Pni P0ni 0 on Gc2

    and ZG

    ec2

    NTs mjpnj dG XG

    ec2

    mjPnijs

    ZGc2

    N0Ts mjp0nj dG

    XG

    ec2

    mjPnijs0

    in which mjPnijs and mjPnijs0 are tangential nodal contact forces on Gc2 :

    Finally, we obtain XZO

    eO0eBTr dO F Rc 8

    in which

    F XZOeO0e

    NT %FF dO XZGesG0es

    NT %pp dG

    is nodal force corresponding to the applied load, and

    Rc XG

    ec2

    mjPnjs XG

    ec2

    mjPnjs0

    is the nodal contact force on the Gc2 : If all contact nodes are in sticking state, Rc will be equal to

    zero and the equation becomes same as conventional FEM. For other contact statues, the

    expressions ofRc will be given in next section.

    2.3. Computational procedure for elasto-plastic contact problems

    By applying the initial stress method usually used for elasto-plastic materials to elasto-plastic

    contact problems, Equation (8) can be expressed in the following incremental form:

    KmDUm DFm DFmep DRmc 9

    where superscript m denotes the application of mth load increment, DU is incremental nodal

    displacement vector, DF is load increment vector, DRc is the increments ofRc; K is the stiffness

    matrix which depends on the contact status (see Appendix A for details), and

    DFmep XZ

    OeO0e

    BTDrme Drmep dO 10

    is the residual force vector in which subscripts e and ep denote elastic and elasto-plastic,

    respectively.

    By uncoupling Equation (9) to contact iteration and elasto-plastic iteration , we obtain:

    Kcmn dU

    mn dRc

    mn 11a

    Kmn dUmn dFep

    mn 11b

    in which subscript n denotes iterative step and prefix d implies the increments of variables after

    an iteration. The corresponding iterative procedure in a load step is illustrated in Figure 4.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 531

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    8/26

    Theoretically, the iteration may begin with either contact iteration or elasto-plastic iteration

    and go on till the contact status of all contact nodes does not change and there is no increase in

    new failure points. In practices, it seems work better that the procedure begins with contact

    iteration.

    The iteration of Equation (11b) is same as normal elasto-plastic iteration procedure. But some

    details concerning iteration of Equation (11a) will be elaborated further.

    2.3.1. dU and Kc. Besides sticking status and slipping, there can exist separating status and

    penetrating status. The former implies that a pair of formerly sticking nodes are separated as the

    deformation develops. The latter implies that formerly apart nodes may penetrate each other in

    the process of contact, since we cannot estimate the magnitude of the displacements in prior.

    Obviously, penetrating status is not physically allowable, so the overlapping part should beforcibly pushed back (see the following text). The four types of contact statuses are illustrated

    in Figure 5 and corresponding criterions are shown in Table III, in which P and U denote nodal

    contact force and displacement, respectively, and CB is equivalent nodal cohesion corresponding

    to cohesion cB:

    To satisfy the displacement conditions on contact boundary, firstly, nodal displacement

    increments are defined under local co-ordinate system, as shown in Figure 6. Secondly, as shown

    in Table IV, independent displacement increments are taken as unknowns, instead of assigning

    displacement increments for every displacements of contact nodal pairs. Hence, the order of

    Equation (11a) is the sum of independent displacement increments and is variational with the

    contact status.

    Noted that dRcmn in Equation (11a) could not be straightaway evaluated with

    dRcmn Rc

    mn1 Rc

    mn 12

    for the order of Rcmn1 may not equal the order of Rc

    mn due to the change of contact status

    from n 1th to nth iteration.

    In order to illustrate how to make Rcmn cooperated with Rc

    mn1; here we assume that a

    nodal pair i change from sticking status into slipping status after n 1th iteration.

    Elasto-plastic iteration

    with equations (11b)

    Convergent?

    contact iterationwith equation (11a)

    Convergent?

    T T

    TT

    F

    F F

    F

    Iteration end

    Contact status changedafter iteration with (11b) ?

    Failure points increasedafter iteration with (11a) ?

    Figure 4. Iterative procedure.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO532

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    9/26

    The equation of n 1th iteration is

    Kcmn1dU

    mn1 dRc

    mn1 13

    d nU

    d nU

    dsU

    d sU

    e

    e

    Figure 6. Nodal displacement increments on contact boundary.

    nP

    sP

    nP

    nP

    0

    0

    sticking slipping

    separating penetrating

    nS

    Figure 5. Contact statuses in computation.

    Table III. Contact status and corresponding criterion.

    Status Criterion

    Sticking Pn50 and jPsj5CB mjPnj

    Slipping Pn50 and Ps5CB mjPnj

    Separating Pn50

    Penetrating Un U0n d

    0n > 0

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 533

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    10/26

    in which

    dUmn1 dUsi T

    before nth iteration, we have:

    Rcmn 0

    T

    Since the slipping nodal pair has two independent displacement increments in tangential

    direction, thus:

    dUmn dUsi dU0si

    T

    Accordingly, dRcmn will be evaluated with

    dRcmn Rc

    mn1 R

    0c

    mn 14

    in which

    Rcmn1 mjPnij mjPnij

    T

    and Rcmn in Equation (12) has been equivalently replaced with

    R0cmn Psi Psi

    T

    as shown in Figure 7.

    Table IV. Displacement increments of a contact nodal pair (for two-dimensional problem).

    Defined displacement Number of independent The displacement conditionStatus increments increments should be satisfied

    Sticking dUndU0n; dUsdU

    0s 2 Un U

    0n d0 0

    Slipping dUndU0n; dUs; dU0s 3 Un U0n d0 0Separating dUn; dU

    0n; dUs; dU

    0s 4 Un U

    0n d050

    Penetrating dUn; dU0n; dUsdU

    0s 3 Un U

    0n d0 0

    d siU

    iP

    s

    d niU

    d siU

    i

    i

    d niU

    d siU siP

    e

    c

    d niU

    e e

    e

    e

    Figure 7. Equivalent transformation of displacements and forces at a sticking nodal pair.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO534

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    11/26

    2.3.2. Rc. Among the above-mentioned four different status, penetrating status is

    a special case. As shown in Figure 8, the overlapping part in normal direction at nodal pair

    j is

    DSnj Unj U0nj d0j 15

    the normal incremental displacements which have to be pushed back are dSnj2 O and

    dS0nj2 O0; respectively,thus

    dSnj dS0nj DSnj 16a

    or

    dSnj ZDSnj dS0nj Z0DSnj Z Z0 1 16b

    In computation, it is difficult to determine the exact values of Z and Z0 beforehand, but that has

    little influence on the final result as long as the load increments are small enough. Usually, if the

    stiffness of the two contact objects are close, we may take Z Z0 12: While if one (such as O0) is

    much stiffer than the other O; we may take Z 1 and Z0 0; so that dSnj DSnj and dS0nj 0:

    Adjusted contactboundary

    njS

    i

    j d njS

    n

    s

    d njS

    cd siR

    cd niR

    e

    e

    Figure 8. Diagram for evaluating dRc at penetrating status.

    Table V. Nodal tractions Rc at nodal pair i:

    O O0

    Status Normal Tangential Normal Tangential

    Sticking Pni Psi Pni Psi

    Slipping Pni mjPnij Pni mjPnij

    Separating 0 0 0 0

    Penetrating dRcin PNP

    j

    kinjn dSnj k0injn

    dS0nj

    dRcis PNP

    j

    kis jn dSnj k0is jn

    dS0nj

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 535

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    12/26

    Sequentially, dRc is evaluated with

    dRcin

    XNP

    j

    kin jn dSnj k0in jn

    dS0nj 17a

    dRcis XNP

    j

    kis jn dSnj k0is jn

    dS0nj 17b

    where subscript n and s denote normal and tangential directions, respectively, and i is all nodes

    (including j itself) of the element to which j belongs. k denotes stiffness component. NP is the

    sum of penetrating node pairs. Finally, nodal contact forces corresponding to different contact

    status are summarized in Table V.

    2.4. Demonstration of the behaviours of interface when taken as a contact problem

    To demonstrcate the effects to treat interface as a contact problem, a simple example similar to

    direct shear test is given in Figure 9. Initially the vertical load s is applied. As the horizontal

    load t increases, relative slip between the contact blocks will occur and this will lead to fast

    increase of the horizontal displacement of upper block (represented with point A on the

    interface). When load t reaches to tmax 2ms 0:6s; the displacement of upper block becomes

    0.5

    0.5interface)( 3.0= Square 1 1

    10000

    0.2

    E

    =

    =A

    0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.50.0

    0.2

    0.4

    0.6

    0.8

    1.0

    1.2

    1.4

    1.6

    1.8 No interface

    3.0=

    2.0=

    1.0=

    Load

    Horizontal displacement (10-4)

    (a)

    (b) (c)

    Figure 9. A simple example to demonstrate the behaviours of interface by taking as contact problem:(a) general description; (b) finite element mesh; and (c) relation between horizontal displacement

    and load under different vertical loads.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO536

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    13/26

    infinite, which implies that all nodes on interface have slipped. It is noted that no special stress

    strain relation is introduced to simulate the mechanical behaviours of interface as in most

    interface models at present and the properties of interface fully depends on cB; m of the interface

    and properties of the deformable bodies.

    3. EXAMPLES

    In this section, several representative examples are presented to assess the validity of the

    method developed in this paper. In most examples, the geotechnical materials are treated as

    2.4 m

    Interface

    Infinite

    P

    M

    3

    o

    26MPa

    0 3

    18kN/m

    30

    0

    E

    .

    c

    =

    =

    =

    =

    =

    (a) (b)

    Figure 10. Rigid strip footing: (a) general description; and (b) finite element mesh (a half).

    -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.20246

    8101214161820222426283032343638

    Analytic solution (P=40kN,e=0)FEM solution (P=40kN,e=0)Analytic solution (P=40kN,e=0.4m)FEM solution (P=40 kN,e=0.4m)

    Pressure(kPa)

    Horizontal coordinate (m)

    Figure 11. Pressure distributions beneath the footing cB 0; m 0:0:

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 537

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    14/26

    ideal elasto-plastic materials and obey MohrCoulomb yield criterion. Otherwise, four-node

    isoparametric element is employed in computation. For material parameters, elastic modulus,

    Poisson ratio, unit weight, internal friction angle and cohesion are denoted as E; n; g; f; c;

    respectively.

    3.1. Contact pressure beneath strip rigid footing

    As shown in Figure 10, a rigid strip footing is applied with line distributed force P or moment

    M: This is a rigid-deformable contact problem. The curves shown in Figure 11 are the contact

    pressure distributions under load P and M; respectively, when the interface is frictionless m

    0:0 and the foundation is elastic. The FEM and analytical solutions show good agreement.

    Figure 12 shows the pressure distributions corresponding to the increase of P when the interface

    is frictional m 0:2 and the foundation is elasto-plastic.

    3.2. Lateral earth pressure on rigid retaining wall

    At first, we compute a rigid retaining wall with frictionless interface m 0; as shown in

    Figure 13. The computational lateral pressuredisplacement curve is shown in Figure 14. Also,

    we can get the lateral pressure distributions on wall back corresponding to different

    displacements of the wall. As an illustration, Figure 15 gives the pressure distributions when

    the displacements are 0.0 and 0.004. We can see that all the computational results show very

    good agreement with analytical solutions [19]. Especially, it is worthwhile to note that if the

    interfaces are ignored (take m 1), the computational results (plotted with the dash line) will

    deviate far away from the correct solution. Therefore, it is critical to simulate interfaces with

    reasonable model in computation.

    The second example is an in situ test for lateral pressure on retaining wall shown in Figures

    16(a) and (b). Figure 17 depicts the computational result of the displacement of the wall and

    -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.20

    50

    100150

    200

    250

    300

    350

    400

    450

    500

    550

    600

    650

    700

    750

    P=80kN/mP=240kN/mP=400kN/mP=560kN/mP=720kN/m

    Pressure(kPa)

    Horizontal coordinate (m)

    Figure 12. Pressure distributions beneath the footing under different loads cB 0; m 0:2:

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO538

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    15/26

    soil, in which relative slip along the interfaces is obvious. Figure 18 shows the lateral

    pressure distributions obtained with in situ test and computation, respectively. We may observe

    the similarity between the test and computation results. Moreover, the maximum pressures

    gained by computation and test are 44.1 and 44:4 kPa; respectively, though the locations

    are different.

    3.3. Behaviours of axially loaded bored piles

    The two piles computed have come from a in situ test studying the behaviours of axially loaded

    bored piles in the Xigeda mudstone , which is a special mudstone distributed in south of Sichuan

    16.7mRigid Retaining Wall

    Interface

    backfill

    Interface

    Interface3

    o

    26MPa

    0 3

    18kN/m

    300

    E

    .

    c

    =

    =

    =

    ==

    (a)

    (b)

    Figure 13. Rigid retaining wall: (a) general description; and (b) finite element mesh.

    -0.010 -0.005 0 .0000

    20

    40

    60

    80

    100

    120

    -0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30

    0

    100

    200

    300

    400

    500

    600

    700

    Towards the backfillaway from the backfill

    Analytic solution

    FEM solution

    Earthpressure(kN)

    Displacement of the wall (m)

    Figure 14. Lateral earth pressure against displacement of the wall.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 539

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    16/26

    province, Peoples Republic of China. The Pile 1 is impact boring pile and the Pile 2 is hand

    digging pile with open end. To study the mudstone exclusively, the overburden around piles was

    replaced with loose fill during the test. The piles were also wrapped with sheet iron to reduce

    the friction between pile and fill. The characterization of the piles and strata are shown in

    Figure 19(a). The elastic modulus E was obtained from an in situ plate loading test, and c; f

    of the mudstone were measured in the lab. The other strength parameters of the interface

    are shown in Table VI, in which m and mR are the coefficient of friction before and afterslip occurs.

    Obviously, this can be computed as an axisymmetric problem. Figure 20 depicts the

    development of relative slip range along the skin of Pile 1 with the increase of load, while

    Figure 21 shows failure zone near the bottom of the pile. In Figure 22, we can see that

    the computational loadsettlement curve of Pile 1 shows good agreement with the test curve on

    the whole. However, when the load exceeds 6000 kN; the development of computational curve is

    smoother than the test curve. In fact, as the load increases, failure surface occurs and grows in

    the mudstone near the base of the pile, which cannot be completely simulated with the ideal

    elasto-plastic model. Also, the result ignoring the interface is given in Figure 22, which is nearly

    linear and has little similarity with the test curve.

    Pile 2 was cast as an open-end pile designed to investigate the skin friction exclusively.

    Unfortunately, the test load did not achieve ultimate load due to the under estimation of thepiles bearing capacity. As shown in Figure 23, the loadsettlement curve of test and

    computation are almost linear before ultimate load. In computation, however, when the load

    approaches 14 300 kN; the iteration becomes divergent, which implies that all nodes on interface

    have slipped and the pile lost support so as to subside in whole.

    5

    4

    3

    2

    1

    0

    -50 -40 -30 -20 -10 0 10 20 30 40

    Analytical solutionEarth pressure at rest

    (FEM solution, ignoring interfaces)

    Earth pressure at rest

    (FEM solution)

    Earth pressure when the displacement

    of the wall is 0.004m towards backfill

    (FEM solution)

    Depth(m)

    Earth pressure (kPa)

    Figure 15. Lateral earth pressure distributions.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO540

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    17/26

    3.4. A shield-driven metro tunnel

    As an example for contact problem with initial gap between the two contact bodies, a shield-

    driven metro tunnel is computed. Usually, the gap between lining and surrounding, which

    mainly arose due to the difference between the external diameter of shield machine and

    that of the lining, can hardly be fully filled. The characterization of the tunnel, strata and

    computational mesh are shown in Figure 24. In this computation, we take the maximum

    gap d0max 25 mm: The loads from buildings, vehicles, etc. are simplified to a 20 kPadistributed load.

    The settlement of ground surface, the internal forces in the lining, etc. can be obtained from

    the computation results. However, we focus our interest on the contact process between the

    lining and the surrounding strata here. In Figure 25, 20, 40, 100% are the percentages that the

    excavation loads were released. The results showed that the contact begins at the top and

    Excavation boundary

    1:0.

    16

    3070mm

    5000mm

    5200mm

    3200mm

    1200mm

    1:

    0.65

    Backfill

    InterfaceInterface

    )3.0( =

    )5.0( =

    Interface (=3.0)

    3

    o

    15MPa

    0 3

    20kN/m

    25

    50kPa

    E

    .

    c

    =

    =

    =

    =

    =

    3

    o

    5MPa

    0 3

    18kN/m

    15

    20kPa

    E

    .

    c

    =

    =

    =

    =

    =

    3

    o

    2 5MPa

    0 35

    18 9kN/m

    5

    17kPa

    E .

    .

    .

    c

    =

    =

    =

    =

    =

    (a) (b)

    (c)

    Figure 16. A retaining wall in engineering: (a) profile; (b) parameters of the soils and interfaces;and (c) finite element mesh.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 541

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    18/26

    bottom of the tunnel, then develops towards the middle, and the lining and the surrounding

    strata become fully contacted in the end.

    3.5. Tunnels in landslide

    The two tunnels are on the ChengduKunming Railway in southwest of Peoples Republic of

    China. The first one was constructed in 1971. Since a heavy rains in 1991, the tunnel began to

    0 10 20 30 40 500

    1

    2

    3

    4

    5

    6

    7

    8

    9FEM solution

    Result of in-situ test

    Dist

    ancetothebottom

    ofthewall(m)

    Horizontal pressure (kPa)

    Figure 18. Horizontal earth pressure on the wall.

    Original configuration

    Deformed configuration of the

    undisturbed soil or displacement

    of the wall

    Deformed configuration of the backfill

    Figure 17. Displacement of the wall and soil.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO542

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    19/26

    deform increasingly that the lining spilt and the track was severely bent and moved 0.51 :5 m

    horizontally. Another tunnel was constructed in 1993, at a distance of 60 m from the old one.

    The similar split occurred after the tunnel was completed in 1994. The subsequent geological

    survey revealed that the tunnels were situated in a landslide with three sliding surfaces, as shownin Figure 26(a).

    The three sliding surfaces and the interfaces between tunnel and stratum are taken as contact

    interfaces in the computation. The final deformation of the landslide and the tunnels is depicted

    in Figure 27, in which the slide of sliding bodies, the movement of old tunnel, and the distortion

    of new tunnel are all salient.

    Earth fill (Formerly boulder or cobble with earth)

    Interface

    Lightly weathered Xigeda mudstone

    kPa7035kN/m5.183.0MPa150 o3 ===== cE

    21m

    Reinforced concrete

    4m

    6m

    2m

    (For pile 1)

    (For pile2)

    Intensively weathered Xigeda mudstone

    Pile

    (a) (b)

    Figure 19. Two test piles: (a) general characterization of the piles and the strata; and (b) finite elementmesh (axisymmetric).

    Table VI. The strength parameters of interface.

    Pile 1 Pile 2

    cB (kPa) m mR c (kPa) m mR

    Earth fill 0 0.2 0.2 0 0.2 0.2Intensively weathered mudstone 100 0.8 0.7 100 0.7 0.45Lightly weathered mudstone 80 1.00 0.8 80 0.9 0.50

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 543

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    20/26

    2000kN 4000kN 6000kN 8000kN 10000kN

    Figure 20. Slip range along Pile 1.

    4000kN 6000kN 8000kN 10000kN 12000kN

    Figure 21. Failure zone near the bottom of Pile 1.

    35

    30

    25

    20

    15

    10

    5

    00 2000 4000 6000 8000 10000 12000

    Load (kN)

    In-situ test

    FEM

    FEM(Ignoring interface)Settlement(mm)

    Figure 22. The loadsettlement curve of Pile 1.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO544

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    21/26

    4. CONCLUSION

    Simulation of interface plays an important role in computation of most geotechnical problems. Theinvestigations showed that no additional constitutive relation for interface needs to be introduced

    when taken as a contact problem. The mechanical behaviours of the interfaces completely depend on

    the shear strength criterion (such as MohrCoulomb law) of interface and the properties of contact

    bodies. Therefore, the required parameters are primarily cB and m (or other strength parameters)

    which are relatively easy to obtain through tests comparing to the interface parameters defined by

    Tunnel

    0.0m

    1.9m

    5.2m

    6.8m

    27.85m

    Weathered Silty Mudstone

    Residual Soil

    Earth Fill

    Silty Mudstone

    Lining

    Excavation Boundary

    Initial Gap

    6250

    mm

    6000mm

    5400mm

    Grout0 maxAlluvial/ Diluvial Soil

    (a) (b) (c)

    Figure 24. Shield-driven metro tunnel: (a) geological section; (b) initial gap between the lining (filling) andthe surrounding; and (c) finite element mesh.

    16

    14

    12

    10

    8

    6

    4

    2

    00 2000 4000 6000 8000 10000 12000 14000 16000

    Load (kN)

    FEM

    In-situ testSettlement(mm)

    Figure 23. The loadsettlement curve of Pile 2.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 545

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    22/26

    The excavation boundary before deformation

    The excavation boundary after deformation

    The lining after deformation

    The excavation boundary before deformation

    The excavation boundary after deformation

    The lining after deformation

    The excavation boundary before deformation

    The excavation boundary after deformation

    The lining after deformation

    (a) (b) (c)

    Figure 25. The contact process between the lining and the surrounding: (a) 20%; (b) 40%; and (c) 100%.

    Figure 26. Tunnels in a landslide: (a) description of the landslide and the tunnels;and (b) finite element mesh.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO546

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    23/26

    other interface constitutive models. Otherwise, the technique developed in the paper make it easier to

    hold the compatibility conditions on interface in the process of computation. Hence we conclude

    that, the method presented in this paper is very practical for the application in engineering problems.

    APPENDIX A: FORMULATION OF STIFFNESS MATRIX FOR

    RIGID-DEFORMABLE CONTACT PROBLEM

    The stiffness matrix for rigid-deformable contact problem consists of elements corresponding to

    deformable body and rigid body and depends on the contact status. We can assume

    Kc KRR KRD

    KDR KDD

    " #A1

    in which subscript R and D denote rigid and deformable, respectively.

    Plane problem were taken as an example to illustrate the procedure of formulating the

    stiffness matrix. The displacement vector of the rigid body is

    V V1 V2 Y T

    in which V1 and V2 are rigid translations in the directions of x1 and x2; respectively, and

    Y is rigid rotation. The displacement of a node on contact boundary Gc is (as shown in

    Figure A1)

    u un

    us

    ( )

    sin a cos a r cosa b

    cos a sin a r sina b

    " # V1V2

    Y

    8>>>:

    9>>=>>;

    An

    As

    " #V AV A2

    0 20 40 60 80 100 120 140 160 180 200

    0

    20

    40

    60

    80

    100

    120

    Old tunnelNew tunnel

    Verticaldistance(m)

    Horizontal distance (m)

    After excavation

    of the new tunnelAfter excavation

    of the old tunnel

    After excavation

    of the new tunnel0.12m

    0.61m

    0.51m

    Figure 27. Deformations of the landslide and the tunnels.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 547

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    24/26

    The contact force P Pn PsT at a contact node will apply force Q Q1 Q2 M

    T on rigid

    body. According to the definition ofA; we can obtain

    Q Qn Qs ATn Pn A

    Ts Ps A

    TP A3

    1. KDDAccording to the definition, the force at node i arising due to displacement at node j is

    Pi Pn

    Ps

    ( )i

    knn kns

    ksn kss

    " #ijun

    us

    ( )j kijn k

    ijs

    un

    us

    ( )j

    kijn T

    kijs T

    " # unus

    ( )j

    k

    ij

    u

    j

    A4

    in which k is the same as the stiffness component in conventional FEM, so we can obtain the

    components ofKDD

    kijDD k

    ij A5

    2. KDR and KRDNotice that the normal displacement un and tangential displacement us on Gc1 ; un on Gc2

    depend on the rigid displacement V; us on Gc2 is independent. Therefore, the nodal force P at

    node i arising due to V is

    Pi XjGc1

    kijuj XjGc2

    kijnujn X

    jGc1

    kijAjV XjGc2

    kijnAjnV

    X

    jGc1

    kijAj X

    jGc2

    kijnAjn

    0@

    1AV A6

    in which S means summing the nodes on Gc1 or Gc2 :

    1Q

    2Q

    M

    1x

    2x

    n

    s

    O

    nPs

    P

    c

    r

    V

    u = A V

    s su = A V

    sP

    n

    n n

    P

    T T

    n n s sP P+ =A A Q

    (a) (b)

    Figure A1. Relations between the displacements (or forces) of rigid body and the displacements (or forces)of a node on the boundary.

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO548

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    25/26

    Finally, we obtain the components in KDR

    kiDR X

    jGc1

    kijAj X

    jGc2

    kijnAjn A7

    By symmetric condition, we have KDR KRD:3. KRRKRR represents the forces acting on rigid body arising from the unit displacement itself. By the

    definition, we can obtain

    KRR X

    iGc1

    XjGc1

    AiTkijAj XiGc1

    XjGc2

    AiTkijnAn

    X

    iGc2

    XjGc1

    AinTkijAj

    XiGc2

    XjGc2

    AinTkijnnAn A8

    In fact, with Table A1, different types of stiffness elements can be easily obtained. However,

    direct application of above formulae in programming is very inconvenient. Therefore,

    equivalently we rewrite, for example

    kijAj knn kns

    ksn kss

    " #ijAjn

    Ajs

    " #

    kijnnAjn k

    ijnsA

    js

    kijsnAjn k

    ijssA

    js

    " #A9

    AiTkij kijnnAjn

    T kijsnAjs

    T kijnsAjn

    T kijssAjs

    T A10

    AiTkijAj kijnnAin

    TAjn kijnsA

    in

    TAjs kijsnA

    is

    TAjn kijssA

    is

    TAjs A11

    Take the evaluation of Equation (A9) as an example. Let subscripts l; m 1; 2 denote normal

    direction n and tangential direction s; respectively, and p 1; 2; 3 denote rigid translations indirections of x1; x2 and rigid rotation, then Equation (A9) may be written as

    keilp X2m1

    kijlmA

    jmp l; m 1; 2; p 1; 2; 3 A12

    Table A1. The components to constitute elements of stiffness matrix.

    D R

    n s On Gc1 On Gc2

    D n kijnn

    kijns

    kijn

    TAj kijnnAj

    n

    s kijsn kijss k

    ijs

    TAj kijsnAjn

    R On Gc1 AiTkijn A

    iTkijs AiTkijAj AiTkijnA

    jn

    On Gc2 kijnnA

    in

    T kijnsAin

    T AinTkijn

    TAj AinTkijnnA

    jn

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    A FINITE ELEMENT APPROACH TO SOLVE CONTACT PROBLEMS 549

  • 8/2/2019 A finite element approach to solve contact problems in geotechnical engineering

    26/26

    Hence the procedure is

    (1) Compute stiffness component kijlm; which is same as the component in conventional FEM.

    (2) Use Equation (A12) to evaluate keilp; which is the contribution of kijlm to Kc:

    (3) Let subscripts L and P; which are corresponding to il and p; respectively, denote the

    location of keilp in global stiffness matrix Kc; then we can assemble keilp to Kc with

    KcLP KcLP keilp A13

    in which implies evaluate, not equal to. The above procedure is corresponding to

    the evaluation ofKDR in Kc: For KDR and KRR; the procedures are similar.

    REFERENCES

    1. Goodman RE, Taylor RL, Brekke TL. A model for the mechanics of jointed rock. Journal of the Soil Mechanics andFoundations Division, Proceedings of American Society of Civil Engineers SM3 (ASCE) 1968; 94:638659.

    2. Zienkiewicz OC, Best B, Dullage C, Stagg K. Analysis of nonlinear problems in rock mechanics with particularreference to jointed rock systems. Proceedings of the Second International Congress on Rock Mechanics , Belgrade,1970.

    3. Clough GW, Duncan JM. Finite element analyses of retaining wall behaviour. Journal of the Soil Mechanics andFoundations Division, Proceedings of American Society of Civil Engineers SM12 (ASCE) 1971; 97:16571673.4. Ghaboussi J, Wilson EL, Isenberg J. Finite element for rock joints and interfaces. Journal of the Soil Mechanics and

    Foundations Division, Proceedings of American Society of Civil Engineers SM10 (ASCE) 1973; 99:833848.5. Desai CS, Zammzn MM, Lightner JG, Sirwardane HJ. Thin layer element for interfaces and joints. International

    Journal for Numerical and Analytical Methods in Geomechanics 1984; 8:1943.6. Wilson EA, Preson B. Finite element analysis of element different displacement. International Journal for Numerical

    Methods in Engineering 1970; 2:387395.7. Francavilla A, Zienkiewicz OC. A note on numerical computation of elastic contact problems. International Journal

    for Numerical Methods in Engineering 1975; 9:913924.8. Okamoto N, Nakazawa M. Finite element incremental contact analysis with various frictional conditions.

    International Journal for Numerical Methods in Engineering 1979; 14:337357.9. Conry TF, Seireg A. A mathematical programming method for design of elastic bodies in contact. Journal of Applied

    Mechanics (ASME) 1971; 2:387392.10. Campos LT, Oden JT, Kikuchi N. A numerical analysis of a class of contact problems with friction in elastostatics.

    Computer Methods in Applied Mechanics and Engineering 1982; 34:821845.

    11. Fischer U, Melosh RJ. Solving discretized contact problems using linear programming. Computers and Structures1987; 25:661664.

    12. Zhong WX, Sun SM. A finite element method for elasto-plastic structure and contact parametric quadraticprogramming. International Journal for Numerical Methods in Engineering 1988; 26:27232738.

    13. Peric D, Owen DRJ. Computational model for 3-d contact problems with friction based on the penalty method.International Journal for Numerical Methods in Engineering 1992; 35:12891309.

    14. Bathe KJ, Chaudhary A. A solution method for planar and axisymmetric contact problems. International Journal forNumerical Methods in Engineering 1985; 21:6588.

    15. Chaudhary AB, Bathe KJ. A solution method for static and dynamic analysis of three-dimensional contact problemswith friction. Computers and Structures 1986; 24:855873.

    16. Heegaard JH, Curnier A. An augmented Lagrangian method for discrete large-slip contact problems. InternationalJournal for Numerical Methods in Engineering 1993; 36:569593.

    17. Katona MG. A simple contact-friction interface element with applications to buried culverts. International Journal for Numerical Methods in Engineering 1983; 7:371384.

    18. Gens A, Carol I, Alonso EE. An interface element formulation for the analysis of soil-reinforcement interaction.Computers and Structures 1989; 7:133151.

    19. Mao JQ. A finite element approach for contact problems and its application in geotechnical engineering.Ph.D. Dissertation, Southwest Jiaotong University, 2002 (in Chinese).

    Copyright # 2005 John Wiley & Sons, Ltd. Int. J. Numer. Anal. Meth. Geomech. 2005; 29:525550

    J. MAO550