12.

Fourier series aided finite element method (FSAFEM)
12.1 Synopsis
As described in Chapter 11, a conventional three dimensional (3D) finite element analysis of a typical nonlinear geotechnical problem is complex and requires a large amount of computer resources. The Fourier series aided finite element method (FSAFEM) is a means of increasing the computational efficiency of the conventional finite element method, for a special class of 3D problems which have a geometry that does not vary in one of the coordinate directions (out of plane direction), but whose material properties and/or boundary conditions do. This efficiency is gained by assuming that the displacements in the geometrical out of plane direction can be represented using a Fourier series and exploiting its orthogonal properties. Two types of FSAFEM exist, the continuous FSAFEM (i.e. CFSAFEM) and the discrete FSAFEM (i.e. DFSAFEM) and they are described in this chapter.

12.2 Introduction
Conventional 3D analyses of nonlinear geotecnical boundary value problems require large amounts of computer resources (Brown and Shie (1990)). A large proportion of these resources are involved in inverting the global stiffness matrix. Consequently, one way of economising on computer resources is to use an efficient method of inverting this matrix. In this respect the use of iterative solution techniques was described in Chapter 11, where it was shown that, for current computer hardware technology, such techniques can result in economies for linear problems, but are unlikely to result in savings for nonlinear problems. However, the situation may improve with future developments in computer hardware. Another way of simplifying 3D analyses is to exploit any geometrical symmetries that exist. One approach that capitalises on such symmetries is the Fourier series aided finite element method (FSAFEM). The theory behind this method and its implementation is the subject of this chapter. Conventionally, the FSAFEM has been applied to problems with an axi-symmetric geometry (but non axi-symmetric loading and/or variation of material properties) and it is the application of the method to such problems that is considered here. However, the method can also be implemented for problems expressed in terms of Cartesian geometries.

Fourier series aided finite element method / 345 All existing formulations of the FSAFEM have been based on linear elastic material behaviour. Nonlinear problems have been analysed with some success (Winnicki and Zienkiewicz (1979), Griffiths and Lane (1990)). In these analyses the nonlinear behaviour was dealt with using a finite element algorithm in which the global stiffness matrix was based on the linear elastic material properties and the nonlinearity was dealt with by modifying the right hand side of the governing finite element equations. In addition, most of the past implementations of the FSAFEM have assumed that the system forces and displacements have a symmetry about the 6 = 0° direction, where 6 is the angular coordinate. This assumption results in a large saving of computer resources required for an analysis and also considerably simplifies the formulation. However, these assumptions are restrictive. Recently a new nonlinear formulation has been developed (Ganendra (1993), Ganendra and Potts (1995)), which allows the stiffness matrix to be updated during an analysis using the nonlinear material properties. It also places no symmetry constraints on the variation of system forces and displacements. However, options can be included to capitalise on any such symmetry if it exists. The chapter begins by presenting the basic theory behind the FSAFEM, and then describes its implementation.

12.3

The continuous Fourier series aided finite element method 12.3.1 Formulation for linear behaviour
The axi-symmetric geometry of the problem domain allows a cylindrical coordinate system to be defined (r-z-6, see Figure 12.1) such that the r-z plane can be discretised using a 2D finite element mesh. Thus the distribution of variables in the r-z plane can be described using nodal values and conventional 2D finite element shape functions. The distribution of variables in the 6 direction can be described using a Fourier series, e.g. the incremental radial displacement, Au, can be written as:
Au = Au°

Figure 12.1 Cylindrical coordinate system (12.1)

Aw2cos2#+ Aw2sin2<9+ ... +Aw/cos/<9+ Aulsinl0+ ...

where Au°, Au1 and Au' are the 0th, Ith order cosine and /th order sine harmonic coefficients of variable Au respectively. Consider the displacements at a point, Au, with components Aw, Av, and Aw in the r, z, and 6 directions respectively. Displacements Aw can be expressed in vector

346 / Finite element analysis in geotechnical engineering: Theory
form in terms of the shape functions of the appropriate element in the 2D finite element mesh, and the Fourier series of the nodal displacements of the element:

N
tsu {Au} = Av Aw
/=o

I

(12.2)

where: Nj is the shape function of the /lh node defined in the element; Uj, Vf and Wf are respectively the /* cosine harmonic coefficient of radial, vertical and circumferential incremental displacement at the /* node; Uj, Vf and WJ are respectively the Ith sine harmonic coefficient of radial, vertical and circumferential incremental displacement at the fh node; n is the number of nodes in the element; L is the order of the harmonic series used to represent displacement and is equal to the highest order harmonic used. The ensuing procedure for formulating the stiffness matrix for the CFSAFEM is undertaken in a manner similar to that for a full 3D analysis. However, as will become evident, the CFSAFEM uncouples the full 3D stiffness matrix, which has a form shown in Equation (12.3), into a series of smaller independent stiffness matrices of the form shown in Equation (12.4). The variables that are solved for in the CFSAFEM are the harmonic coefficients of incremental displacement at each node in the 2D mesh, i.e. u\, Vj , WJ , uj, V/ , w} .

Arf,

A/?,

Kr.

Arf,.

Afl,

(12.3)

5) Thus in the problem domain defined above. are defined in cylindrical coordinates as: d(Au) dr d(Av) dz Au 1 d(Aw) r ' r dO d(Au) d(Av) dz dr 1 d(Au) d(Aw) t Aw r dO ' dr 1 a(Av) d(Aw) r d6 ' dz Aer ^ Asz {Ae} = Aee Aerz Aere As (12. the incremental strains at a point can be expressed in terms of Fourier series harmonic coefficients of incremental nodal displacements and element shape functions: dr dz —L\ u'cosW + UJsinI0 ^-U Ujcosie + UJsinW dz [ ^-\-W/sinl0+Wjlcosl0 ^-\ dr v/cosie + v/si IN.4} 0 0o[ ] o 0 0 0 [KL] AdLl * . AdL** ARL~l * . -U sinl0 + U cosl0 ^±\ W/cosl0+w/sinl0 (12. Ad1** AR1 * . AdL~1** AdL * . ARL~]** ARL * . ARL** The incremental strains at a point.Fourier series aided finite element method / 347 0 [Kl 0 0 0 0 0 0 0 00 0 00 0 0 0 0 0 0 0 \Kl] 00 0 00 0 00 0 00 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Ad° Ad1 * . assuming a compression positive sign convention. AR1** (12. AR1** 0 0 0 0 0 0 0 0 Ad1* .6) . Ad1** AR° AR1 * .

parallel symmetry terms and orthogonal symmetry terms. [B]. sin/<9 n L u -Z Z i = l /=0 dz dr -w ^ L . Rearranging Equation (12. Conversely. code Asrz Aere / = 1 /=0 Z C 0 S /6> dr dN i N i\ • m dr r ) IN. Note that the cosine coefficient of circumferential incremental displacement is expressed with a sign change. Thus the strains can be written as: (12.7) 8r -sinlO 0 dz 0 0 '-sinie 0 IN. The resulting expressions can then be divided into two parts. Parallel symmetry terms consist of cosine harmonic coefficients of radial and vertical incremental displacements and sine harmonic coefficients of circumferential incremental displacements. orthogonal symmetry terms consist of sine harmonic coefficients of radial and vertical incremental displacements and cosine harmonic coefficients of circumferential incremental displacements.8) .= —V L dr Asz Ase 0 0 0 code IN.cos/(9 ar ry in illLcos/(9 r 0 lN —-codO r i -—LCOs/<9 az Splitting the equation into two parts in this manner results in the [B] matrices for each part having a similar form.i ^ . sin/<9 dz -sin/6> (12. The dashed lines in the matrices separate the sine and cosine terms and split the [B] matrix into a top and bottom section.: dz n .6) in this manner gives: -cos/<9 dNj_ AS.348 / Finite element analysis in geotechnical engineering: Theory These incremental strains can be rewritten as the sum of the product of a series of strain matrices. and vectors of harmonic coefficients of incremental displacements.

[^2i]5 an< i L ^ L ^ accordance with the subdivision of the [B] matrix into two submatrices [B\] and [B2]. [Du]. i. dz and —i Hi.> [51^.]sin/6>" -[B2lj]cosl0 The [D] matrix for a material relates the incremental stress vector {Ac} to the incremental strain vector {As}.11) It can be divided into four submatrices. dz o dr w! (12. dr dz r i and Note: {Ad/*} and {A<//**} are the parallel and orthogonal 7th order harmonic coefficients of incremental displacement for the fi node respectively. [Dn].{Acini -[D]- (12 10 . and [B1 /] and [B2/] are the top and bottom sections of the Ith order harmonic [B] matrix for the ith node respectively.Fourier series aided finite element method / 349 where: BN. This can be written in matrix form as: .e: {A<7} = [D] {As} (12.9) r 0 IN-. The incremental internal work done can now be written as: AW = J |{A^} T {Acr} <\6 r&area -71 = J ]{Au}T[Bf[D][B] {AM} d0 r darea -\)t\i 7=1 '=0 k .

350 / Finite element analysis in geotechnical engineering: Theory
A<7,.

As,. Asz Aee As,r

Aa: Acre
ACT,,

(12.12)

[Ail
As,a

Thus the equation for incremental internal work can be rewritten as:
AW = |

[zv rn ^ ^
[Bl}]sh [52f]cos^<9j [D22]\ h [B2'j]sin!0
d'j }

rJ2J 2
k=0

-[B2?]cosk0

[D2l][D22]\

to

{Ad'j

r dOdarea

(12.13) When integrating with respect to 6, a large number of terms in the stiffness matrix become zero due to the orthogonal properties of the Fourier series: J sin£<9 cos/6> &6 = 0 for all k and /;
-71

fsin£<9 sin/<9d<9
-71

= 0 i f k * /, = % i f k=I* =nifk

0, = 0 ifA: = / = O ; =2nifk =l =0 .

(12.14)

JcosA:6> cos/6» &6 = 0 ifk*I,

= l*0,

Thus the incremental internal work done can now be simplified to:

Fourier series aided finite element method / 351

} r darea -2{Arf / 0 *} T [ J Bl / 0 ] T [ J D 12 ][^.]{A^ / 0 "}- 2{Arff*} T [ J B2, 0 ] T [/) 21 ][51^]{Arff} + +

£

} + {Adj } [52,] [Z>21][j?l ]{Atf

}

I } r darea

(12.15) The stiffness matrix has been uncoupled into L+l smaller stiffness matrices of the form shown in Equation (12.4), with one independent set of equations for each harmonic order. This form of uncoupling is called harmonic uncoupling. However, for each set of equations associated with a particular harmonic order the parallel symmetry displacements, {Ad/*}, and orthogonal symmetry displacements, {Ad/**}, are coupled in the manner shown below:
'[KlY
l op [Kl] ]

[Kl]po

{Ad1"} {Ad1**}

(12.16)

The terms above the dashed line in Equation (12.15) give the independent diagonal terms [Kl]p and [K'Y, and those below give the cross coupling terms [K!]op and [K'Y°- It is noted that for materials with a [D] matrix which has zero off diagonal submatrices, [Z>12] and [D2l], the cross coupling terms disappear and the stiffness matrix reduces to the following form:

In addition, it may be noted that [K'Y = [KlY- In this case the {Ad/*} and {Ad/**} terms for each harmonic order can be solved independently, using the same stiffness matrix. This form of uncoupling is called symmetric uncoupling. The symmetrically uncoupled stiffness matrix for the Ith order harmonic is:
<]T[D22][B2l.]r darea (12.18)

KHTi^rfto)

(12I7)

where:

A= 2if/=0; A=l if/*0.

The applied incremental loads are formulated as harmonic coefficients of incremental nodal force using the equations in Appendices XII. 1 to XII.4. These harmonic coefficients of nodal force are expressed as vectors {AR1*} and

352 / Finite element analysis in geotechnical engineering: Theory
T

{AR '*} = I R'r Rlz Rle \

is the vector of parallel Ith order harmonic coefficients of incremental force;

{ARl**} = <Rlr Rlz -RQ\ where:

is the vector of orthogonal 7th order harmonic coefficients of incremental force;

Rj. , Rlz and Rl9 are the Ith cosine harmonic coefficient of radial, vertical and circumferential incremental force respectively; Rlr , Rlz and Rle are the Ith sine harmonic coefficient of radial, vertical and circumferential incremental force respectively. Thus for each harmonic order two sets of system equations can be written:

} = [Kl]° { A d 1 } The displacements can be solved for by inverting the stiffness matrix in a similar manner to the conventional finite element method. An important feature of this linear elastic formulation is the harmonic uncoupling. A consequence of this uncoupling is that for any harmonic order the solution coefficients of displacements are only non-zero if the applied coefficients of load for the same order are non-zero. Thus the number of harmonics required for an analysis is equal to the number of harmonics required to represent the boundary conditions.

2
1

7

;J

(12.19)

12.3.2 Symmetrical loading conditions
A symmetrical function of #,/.(#), has the property that: /v(#) = / s ( - 0 ) (12-20)

and the Fourier series representation of such a function would only consist of the zeroth (i.e. 0th) and cosine harmonic terms. An asymmetrical function of8,fm(ff), has the property that: fm(O) = -fm(-O) (12.21)

and the Fourier series representation of such a function would only consist of sine harmonic terms. The linear CFSAFEM formulation has divided the solution displacements, {Ad}, into two parts, parallel symmetry displacements, {Ait}, and orthogonal symmetry displacements, {Act*}. The parallel symmetry displacements consist of symmetrical displacements in the r and z coordinate directions and asymmetrical displacements in the # direction. Conversely, orthogonal symmetry displacements consist of asymmetrical displacements in the r and z coordinate directions and

are separated in a similar manner into parallel loads.Fourier series aided finite element method / 353 symmetrical displacements in the 6 direction. or orthogonal symmetry problems (Griffiths and Lane (1990)). The applied loads. then a non-symmetrical . {A/?}.2: Load components on pile Many boundary value problems have a symmetry about the 6 = 0° direction. These boundary conditions can be represented using parallel symmetry.2a-c. and orthogonal loads. For instance. Py = Fx is lateral force applied to the pile = 2nrP b) Lateral load on pile e=o° PZ=P'cosQ r ^ ^ e=o° My = nrP' Pz is the applied vertical force/unit length My is the applied turning moment c) Moment load on pile Figure 12. {AR"}. {A/?*}. see Figure 12. Hence. If the applied loads in a problem do not satisfy any symmetry about the 6 = 0° direction. the loading conditions applied to a pile may consist of a combination of axial loading. such that the imposed boundary conditions consist of purely parallel symmetry or orthogonal symmetry terms. lateral loading and turning moment about an axis perpendicular to the direction of lateral loading. e=o° Pz is the applied vertical force/unit length F2 is axial force applied to the pile = 2nrPz r is the radial coordinate of the node a) Axial load on pile Pr=PcosQ Pe=-PsinQ PX=P . if the direction of lateral loading is parallel with the 6 = 0° direction. all previous implementations of the CFSAFEM have been constrained to analyse either parallel symmetry conditions (Winnicki and Zienkiewicz (1979)).

This is akin to using an elastic stiffness matrix to solve an elasto-plastic problem. in the 6 direction. The tangent stiffness approach cannot be employed because it involves the use of the elasto-plastic stiffness. i. consisting of both parallel and orthogonal symmetry terms. Though this may have been valid for the particular cases analysed.354 / Finite element analysis in geotechnical engineering: Theory formulation is required.3 Existing formulations for nonlinear behaviour All past implementations of the CFSAFEM have used the above formulation for linear elastic behaviour with a constant material stiffness. this method can be problematic for highly nonlinear constitutive models. with non-zero [Dn] and [D2l]. However. only the visco-plastic and MNR approaches can be used in this way. 12. then the resulting solution displacements will also satisfy parallel or orthogonal symmetry. as noted in Chapter 9. with the stiffness matrix calculated using only the elastic part of the constitutive matrix. its use for a material which does not satisfy this criterion is not valid. This results in large savings in computer resources required for an analysis and significantly simplifies the solution algorithms. There is some scepticism about the validity of the above approach for nonlinear problems. Considering the three solution strategies described and compared in Chapter 9. However. nonlinear analyses have been undertaken using the linear elastic formulation. see Figure 12. in combination with a solution strategy which continually adjusts the right hand side of the governing finite element equations to cater for material nonlinearity. The viscoplastic approach was used by Winnicki and Zienkiewicz (1979) and Griffiths and Lane (1990). Similarly. It is argued that. orthogonal symmetry can represent loads in a direction perpendicular to the 6 = 0° direction. [D] matrix. Analyses using only parallel or orthogonal symmetry terms benefit from the reduced number of solution coefficients of displacements that have to be solved for. The term parallel symmetry stems from the fact that it can be used to represent a load in a direction parallel to the 6 = 0° direction. No formulation has been developed for nonlinear material behaviour with a variable [D] matrix in the 6 direction.3. The previous implementations of the CFSAFEM have assumed that if the applied loading conditions can be represented using either parallel or orthogonal symmetry. respectively. since the formulation of the CFSAFEM assumes that the [D] matrix is constant in the 6 direction. A better alternative is the use of the MNR approach.2b. this assumption has been shown to be true only for materials with a [D] matrix which has zero off diagonal terms [Dn] and [D2l]. The popular practice of using purely parallel or purely orthogonal symmetry terms in an analysis requires that the system stiffness . the analysis of an anisotropic elastic material.e an isotropic linear elastic material. In the linear elastic formulation presented above. it is not true in general. irrespective of any symmetry in the imposed loads. Due to the coupling of parallel and orthogonal terms. would require the full non-symmetrical formulation.

. is used to formulate a system stiffness matrix as described by Equation (12. Thus the number of harmonics used in an analysis is an important parameter which is based on a trade off between solution accuracy and computer resources. The anomaly is overcome by allowing the corrective loads.. but is also dependent on the material nonlinearity. An important feature of nonlinear CFSAFEM analyses is that harmonic uncoupling no longer exists. Separating the cosine and sine harmonics of the [D] matrix allows Equation (12. applied by the procedure for nonlinear finite element analysis. in terms of a series of products of harmonic coefficients of displacement and harmonic [B] matrices. A new nonlinear formulation is proposed for the CFSAFEM which incorporates a variation of [D] in the 0 direction.3. [D2l] and [Z>22] terms. 3 . (12. L. However. [Df] and [Dl] are matrices containing the 0th. The number of harmonics required to represent the solution displacements is not just a function of the number of harmonics required to represent the boundary conditions..].1. the [D] matrix is no longer a constant. 1 2 . / / . In general. but varies with 6. Ith order cosine and Ith order sine harmonic coefficients of the components of the [D] matrix respectively. as noted above. this assumption is not valid for nonlinear material behaviour.13). These harmonic [D] matrices can also be split into 4 parts representing [/>.. need not be the same as the number used to represent displacements.13) to be rewritten as: . Increasing the number of harmonics used in an analysis increases both the amount of computer resources used and the solution accuracy. there has never been a rational for it. Past practice has been to allow a greater number of harmonics of solution displacement than is required to represent the boundary conditions. [Dl2]. This variation can be represented with a Fourier series for each component of the [D] matrix: [D] = [D°] + [Dl]cos0+[Dl]sin0+[D2]cos20+[D2]sin20+ [Z> ]cos/6>+[Z) ]sin/<9+ . This feature cannot be reproduced with the linear elastic formulation which. This criterion is not satisfied in general for elasto-plastic materials and thus questions the validity of the above practice for nonlinear analyses. The linear elastic CFSAFEM formulation of strain. since the [D] matrix is now stress history dependent and the stresses vary in the 6 direction. to have harmonics other than those associated with the applied boundary conditions.Fourier series aided finite element method / 355 matrix is symmetrically uncoupled. 4 N e w formulation for nonlinear behaviour The linear formulation for the CFSAFEM assumes that the [D] matrix is constant in the 6 direction. Though this has given reasonable results. The number of harmonics used to represent the [D] matrix.22) where [D°]. as discussed in Section 12. only yields non-zero harmonics of displacement if a load of the same harmonic order is applied.. The criterion for this is that the off diagonal terms in the [D] matrix are zero. M.

for the nonlinear formulation the corresponding integration is of the product of three Fourier series and a new set of solutions has to be derived.-} {Ad1"} r darea | 1 •LJ'yj | I -D-i -J I Y.5.]cos/^ [B2lj]sinl0 [Bl!j]sinl0 -[B2*j]cosl0 -[B2j]cosl0 {Arfj } k=0 L k=0 [B\f]cosk0 [B2f]sink0 [Blf]sink0 -[B2f]cosk0 {Ad1**} {Ad1. 10 J r ni'i 1 i* ^ • 1 -j o\ r L\ [DU' ] [Din ] A/ o kJ (12. Thus the incremental internal work done can be written as: /= 1 . see Appendix XII.*} [D%\]cosm0[D2z]cosm0 [Blf]cosk0] M [B2f]sink0j w=c [D 1\]sinm0[D l]sinm0 £<\[B2'j]sinI0\ 2 2 [B\f]sink0] M [Dfl]sinm0[D?2]sinm0 [D™]]smm0[D2 2 [B2f]sink0 [B\*]sink0 ~[B2f]cosk0 [D"\]sinm0[D"2]sinm0 [D^\]smm0[D22]sinm0 } [Bl'j)cosl0] t ]sinm0 [Df[]sinm0[D?2]sinm0 k=0 £ -[J?2'.} {Ad1. the dO integral is performed and a large number of terms in the stiffness matrix are zero due to the orthogonal properties of the Fourier series.] {Arff} [B2'j] \BI'J]Z [B2lj}_ 1 XJ 1 • 1 k=0 £ £ /=o L r nk J ~\()P r T\k J ~\®P 1 -^^1 1 1 1 -^H 9 1 1 D^l 1 -^^11' I \. However. For the linear formulation this integration was solved using the standard solutions for the integration of the product of two Fourier series.]cos/0j (12. 1\ J {Ad1./=1 k=0 [Bit]' [B2f] T [Blf]" [Blf]' [B2t]_ T k [B2>]_ \[Dtfy [D^f i=o\_[D2{'Y [D22Y [Bl'.32) r d0darea As with the linear formulation.356 / Finite element analysis in geotechnical engineering: Theory [Dj[]cosm0[D?2]cosm0 -%i=\j=\k = O [B\!j]CQSl0 I [B2?]sink0] [D!?l]cosm0[D!?2]cosm0 1=0 [B2 j]sinl0 [Dj[]cosm0lD?2]cosm0 [Ul'.24) .

The nonlinear CFSAFEM stiffness matrix does not exhibit harmonic uncoupling. similar to that for a full 3D analysis. [Dkl]po and [Dk'!]° are the [D] matrices resulting from the d6 integration that should be used to form the stiffness matrix relating the kih harmonic coefficients of load with the Ith harmonic coefficients of displacement.ifk.l<\. is . The components of these matrices can be evaluated from the following equations: [Dkir = [Dklf = [DkJ]"" = (12. = V 2 otherwise. is + i f k .e. i.if A . is. These [D] matrices have been split into four parts representing [Du].25) where: a=lifk=l. [D]2]./ > 0.Fourier series aided finite element method / 357 where: [DkI]p. [DkJY is the [D] matrix that relates parallel loads to parallel displacements./ > 0./<1. This explains the important feature observed in nonlinear problems that solution displacements have harmonic terms with orders different from that of the applied loads. ± i s + i f A . 0=lifk = l = O. [Dk!]op is the [D] matrix that relates orthogonal loads to parallel displacements. [DkJY° is the [D] matrix that relates parallel loads to orthogonal displacements. [DkJY is the [D] matrix that relates orthogonal loads to orthogonal displacements. The harmonically coupled stiffness matrix from this formulation is very large and a large amount of computer resources. [D2]] and [D22] terms. the zero terms in Equation (12. = V2 otherwise. + is .4) are now non-zero. [DkI]op.

as discussed in Chapter 9. Since the correct elasto-plastic stiffness matrix is used to solve each set of harmonic system equations. the proposed partial nonlinear formulation is expected to require less computer resources for its correction process than the simple linear elastic formulation. The tangent stiffness solution strategy could be implemented using the full nonlinear CFSAFEM formulation. az0). and the [D]2] and [D2]] parts are asymmetric functions of 6. for a purely orthogonal symmetry analysis the [D] matrix criterion must be satisfied for an orthogonal symmetry stress and strain state. In a parallel symmetry analysis the stresses associated with the top part of the [B] matrix. i. . For symmetrical uncoupling to be valid the [D]op and [D]po matrices must be zero. thus allowing the system equations for each harmonic order to be solved individually. oz. The use of the linear CFSAFEM formulation to solve strongly nonlinear problems also requires a large amount of computer resources. An orthogonal symmetry stress and strain state exists when the stresses and strains associated with [Bl] are asymmetrical functions and those associated with [B2] are symmetrical functions of 0. i. since the elastic [D] matrix used in the formulation is very different from the true [Dep] matrix for the material.24). Similarly. The error associated with this omission is corrected using the nonlinear solution strategy in a similar manner to the error associated with the linear formulation. and the number of harmonics used to represent the applied loads. or only sine harmonics. The nonlinear formulation is now able to yield a more rational criterion for the symmetrical uncoupling of the system equations. the k * I terms in Equation (12.e. are symmetrical functions of #. while the stresses associated with the bottom part. It should be noted that the linear CFSAFEM formulation is a particular case of the nonlinear CFSAFEM where the [D] matrix does not vary with 6.. The strains satisfy the same symmetrical conditions. Thus a material can only be legitimately analysed using purely parallel symmetry terms if it satisfies the [D] matrix criterion while subjected to a stress and strain state associated with parallel symmetry. the solution displacements and the [D] matrices.e.358 / Finite element analysis in geotechnical engineering: Theory required to invert it. The resulting [D]op and [D]po would then be zero. but the accuracy of the solution would be strongly influenced by both the size of the solution increment. are asymmetrical functions of 6.. [B2] (or0. The four submatrices associated with each of these two matrices contain either only the 0th and cosine harmonics. Thus the criterion the [D] matrix must satisfy for symmetrical uncoupling is that the [Du] and [D22] parts of the [D] matrix are symmetric functions of 0. [Bl] (<r. A compromise is to use the nonlinear formulation and omit the harmonically coupled terms. and a large amount of resources is required for the iterative correction process to account for this error. a0 and arz). only the 0th harmonic [D] matrix exists.

5 Formulation for interface elements A formulation for zero thickness Z. v.3. see Section 3. using u. The incremental global displacements on each side of an interface element can be represented using isoparametric shape functions and a Fourier series in the 6 direction. u displacements for the interface elements have to be defined such that: Aw7 is the incremental local tangential Figure 12. see Section 3. As is the in plane normal strain. The relationship between incremental global and local displacements can be expressed in matrix form as: A«.26) 0 0 1 Aw For any point on the interface there are 2 sets of displacements.3. n is the number of nodes on each side of the interface. Ay0 is the circumferential shear strain. L is the order of the harmonic series used to represent displacement. The interface elements have three components of incremental strain: Ayp is the in plane shear strain. is the incremental local circumferential tangential displacement. Local r.6. UJ is the Ith sine harmonic coefficient of radial displacement at the fh node. The global displacements are defined in the same manner as for the solid elements. Aw.6. and w displacement components. = -sina cos or 0 i Av (12. COStf sin a 0 Aw Av. The coordinate f f system for the isoparametric six and four noded element is presented in Figure 12.27) where: Nj is the shape function of the fh node defined in the element. each set describing the displacement on one side of the interface. top and bottom displacements. . Uj is the Ith cosine harmonic coefficient of radial displacement at the /lh node.3: Coordinate system for interface elements displacement in plane r-z\ Av/ is the incremental local normal displacement in plane r-z\ Aw. V isoparametric interface elements for the CFSAFEM has been developed based on ^^the 2D formulation of Day and Potts / ^ (1994).Fourier series aided finite element method / 359 12. in a similar manner to the solid element formulation: Au=±Ni[£ (Uj cos/0+£// sin/0)] (12.

The interface element stresses and strains have been formulated in such a manner that the resulting equation for the internal incremental work has the same form as that for the solid element: . [Blil is the top part of the strain matrix for the /lh node.27) for the incremental global displacements gives: Kl +cosor +sina 0 ±sina Tcosa 0 0 0 +1 UJGOsW+UlsmW £*.29) where: T is . the elasticity matrix [D] for the interface element can be split into four parts: ' ACT P As (12.bot .{ V/cos!0+V/sinI0 W cosI0+wSsml0 l (12.for a bottom node.360 / Finite element analysis in geotechnical engineering: Theory These can be related to the incremental global and local displacements as: K As [ = Au™ Av.]=N.'0" cos a sina ol Aub bot -sina cos a 0 <Av bot 1 0 0 Aw - Autop Av top Awtop (12. ACT is the incremental in plane normal stress.Av.30) where: and {Ad/**} are respectively the parallel and orthogonal Ph order harmonic coefficients of incremental displacement for the /lh node. [B2j] is the bottom part of the strain matrix for the fh node.28) We Using Equation (12. This can be written as: £ /To -[B2i ] cos/6 {A<*} (12.bot top .32) ATa where: Ar^ is the incremental in plane shear stress. .AM.31) Similar to solid elements. so that: [Bl. ± is + for a top node and is . +cosa +sina 0 ±sma 0 0 +1] (12. AT0 is the incremental circumferential shear stress.for a top node and is + for a bottom node.

5): . 12.. the system equations for each harmonic of a problem domain can be solved independently.]cos*0 h [[B2j]smw\ { This can be simplified to give: AW = * \-lB2j\cosW r&6 darea (12.34) Since both interface and solid element equations are harmonically uncoupled. The linear elastic interface element formulation can be extended for nonlinear behaviour using the same approach as for the solid element. The [D] matrix is represented using a Fourier series and the equation for virtual work now yields a harmonically coupled set of system equations. For linear elastic material behaviour the principle of effective stress gives.4 for the solid elements. The [D] matrix pertinent for each harmonic combination of equations is the same as those presented in Section 12. such that there is little dissipation of excess pore water pressure.]T[i)22][B2. symmetrical uncoupling is also valid.6 Bulk pore fluid compressibility Saturated clays are two phase materials consisting of a compressible solid phase. This formulation is now extended for use in CFSAFEM analyses. even if it consists of both types of elements. see Section 3..3.2) and (3.* ' j &W= *f0 [[Bl^sinkOl [Bl7]smk0 |-[J?2.7=1 -{^)\B\i]T[Dn][B2j]{Ad1.4. Similarly. see Equations (3.3.33) n\±± = 1. the soil skeleton. and a highly incompressible fluid phase .the pore water.7=1 2{Ad?*}T[Bll]T[Du][Blj]{Ad?} +2{Adr}T[B2i]T[D22][B2J]{Ad?*} {Arfr}T[B2. if the [Dn] and [D2l] terms in both solid and interface elements are zero.Fourier series aided finite element method / 361 . This type of behaviour can be modelled in a conventional finite element analysis by specifying a bulk pore fluid compressibility.} + r darea (12.]{Arff} rdarea '=1. Undrained behaviour is assumed if these clays are loaded quickly.

6). For the Ith harmonic it is: t ± 1 [tfljftol} K {rfl}T[Bllf] r darea (12. 1.362 / Finite element analysis in geotechnical engineering: Theory } + [Df]{Ae} [Df]{A£} = {rj}Ke Aev (12.36) ^ T T Asv r (12. Ke is the equivalent bulk modulus of the pore fluid and Aev is the incremental volumetric strain. be written as: (12.'](cos/0 { < } + sin/6> {A<//**}) /=1 1=0 T .36) Noting Equation (3.35) (12. .40) ' I t {?7l}T[B^](cosk0 {M**}+ sinkO {Ad)"}) r d6darea Carrying out the d6 integral gives: AWf=±± J2TI{A<}T[51. AW. the work done by the pore fluid. which was calculated in Sections 12. AWf.42) '•=1 7 = 1 In the above constant Ke formulation.0). Using Equation (12.1 and 123 A. 1. the second term in the above equation can be written as: where {J/}T:={ 1 1 1 0 0 0 }. Therefore only the second integral.8).38) The first integral is the work done by the soil skeleton. the pore fluid pressure is a symmetrical function in a parallel analysis and an asymmetrical function in an orthogonal analysis.3.i \T where {//1} = ( 1.1 and 12.41) Ke {TtiyiBl'jUAdf}} r darea This equation gives the contributions to the global stiffness matrix associated with the pore fluid compressibility.39) =\)±t j] k0 j=] k=0 (cos/£ {A<}+ sin/0 {M^fYBX1^{rjl} Ke (12. Combining Equations (12. the incremental internal work done. has to be evaluated. Aev can be expressed as: [2H. can be expressed as: sv dVol (12.0] ( (12.3 A.37) Accordingly.3. The fluid compressibility terms are both harmonically and symmetrically uncoupled and their contributions to the parallel and orthogonal symmetry stiffness matrices are the same. From here AWf can. The system equations are obtained by adding these fluid compressibility terms to the terms associated with the soil skeleton stiffness calculated in Sections 12.35) and (12.

There are two options for defining Ke: to assign it a constant value. Ith order cosine and /th order sine harmonic coefficients ofKe\ M is the number of harmonics used to represent Ke and need not be the same as the number used to represent displacements. the option for defining Ke as a multiple of the bulk modulus of the soil skeleton is preferred. It must be large enough to inhibit volumetric strains. as shown in Appendix XII. p = 1 if k = I = 0.£ > 0 . From there: . but small enough to avoid ill conditioning problems associated with a very large number in the stiffness matrix.Fourier series aided finite element method / 363 In an undrained material Ke is prescribed a very large value. because these difficulties are avoided. However. = 1.43) e where: K°e . Accordingly. In the former option care must be taken in assigning the value of Ke.k<\./=1 1-0 k=0 ni k *=o B\k] {dk"}r<\area (12. i s . Substituting this into the equation for A Wf gives: *Wj• = J J t t (cos/<9 {M/*} +sin/0 -n i = l /=0 {MD^Bl^iTjl}(12. Finding a suitable value for Ke is particularly difficult when the bulk modulus of the soil skeleton changes significantly during the course of an analysis. . A typical value for Ke is one hundred to one thousand times the bulk stiffness of the soil skeleton. such that the volumetric strains are inhibited.5.44) (K + Z K cosl0 + Kle sin/0)/=i 1 t {7l\}T[B\kj}{coske {M?}+&mkO {Ad***})rd0 darea . is + i f / .i f / . or to define it as a multiple of the bulk stiffness of the soil skeleton.4.45) where: a = 1 if k = /. + is . Kle and Kle are the 0th.k<\. considering the Fourier series formulation. see Section 3.7=1 k=0 The &6 integral is carried using the solutions for a triple Fourier series. Ke could now vary in the 6 direction and would have to be expressed as a Fourier series: (12. L. = V2 otherwise. =V2 otherwise. ± i s + i f / .i f / .k> 0.

Section 12. for example when partial drainage occurs during a loading stage. The pore fluid pressures in the problem domain are described by the element pore fluid shape functions and the Fourier series of the nodal pore fluid pressures of the element: Pf = S Npi ( I plfi codO+plf. can deal with either drained soil behaviour. Accounting for such behaviour in the CFSAFEM was described in the previous section (i.364 / Finite element analysis in geotechnical engineering: Theory In general.3. The hydraulic gradient.4. 12. followed by long term consolidation. sin/0 i=\ 1=0 (12.46) where: T V is the element pore fluid shape function for the fh node. The finite element theory behind such a coupled approach was presented in Chapter 10. To account for such behaviour the equations governing the flow of pore fluid and the mechanical behaviour of the soil must be combined. The criterion for symmetric uncoupling is that Ke is a symmetrical function.e. n. m is the number of pore fluid pressure nodes in the element which is not necessarily equal to the number of displacement nodes in the element.3. In this section this theory is extended for use with the CFSAFEM. where full pore fluid pressure dissipation occurs. p'fj and p!fi are respectively the Ith cosine and sine coefficients of pore fluid pressure for the ?h node.6). there is neither harmonic nor symmetric uncoupling. + \ dr dP/ dz dp/ rd6 m L dr P' -cosld -sin/6> -sin/0 dz -lN P' dz s'mW -^-cosW -ft •f (12. {V/i}. can be defined as: dpf dNn dr dNni •pf. as described in Section 3. As noted in Chapter 3. The latter behaviour is achieved by introducing the effective bulk compressibility of the pore fluid. described in Chapter 2. but subsequently the theory is extended to account for soils which have variable permeability. only soil with a constant permeability is considered. Often soil behaviour cannot be simplified to being either fully drained or undrained.47) .7 Formulation for coupled consolidation The behaviour of saturated soils under any loading condition is strongly influenced by the rate at which the generated pore pressures are able to dissipate within the soil mass. the conventional finite element theory. or undrained behaviour where no dissipation occurs. Initially.

see Equation (12.53) Ap)} sin&<9) r dOdarea /=1 k=0 .Fourier series aided finite element method / 365 where: yf is the bulk unit weight of the pore fluid. {iG} is the vector parallel to gravity.48) The incremental strain vector {AE} can be expressed in terms of the Fourier series coefficients of incremental displacement as in Equation (12. The first integral in Equation (12.49) Thus the incremental volumetric strain can be expressed as: Asv = t t tol}T[*l.38). where \dh) dr dh_ dz_ ~d6 dr dz and dh and The incremental volumetric strain at any point is defined as: A£v = {r]}T{A£} (12.-] (cos/0 {A<//*}+ sin/6> {A<*}) /=1 /=0 (12.51) W (12.52) where [k] is the matrix of permeabilities and Q represents any sources and/or sinks. see Chapter 10.8): [Ul.„ (12. The second integral can be expressed in terms of the incremental nodal pore fluid pressure and displacement Fourier series: \Asv Apf dVol =j]±t (cos/0 { < } + sin/0 {AdDYiBl'f {rj\} • (12.50) The two governing equations are the equilibrium equation: | {A s}1 {A a} dVol +1 Asv Apf dVol = external work done and the continuity equation: (12.']sin/0 .51) is the same as that for the incremental internal work done by a solid element without consolidation. Q is expressed as a Fourier series and the parallel and orthogonal components are Q!* = Ql and Q1** = Q1 respectively.

Carrying out the d6 integral and using the orthogonal properties of Fourier series gives: \AsvApf dVol }T[/?l.52) can be rewritten as: )T [k] {Vh}dVol = .f. = n\r [Biff {J/1} iV otherwise.°]T{/7l} Npj Ap°£ +i{Ad!T[Blii]T{T7l} 1=1 Np} A^ (12. These equations are both symmetrically and harmonically uncoupled. This equation gives the terms that relate the applied incremental loads to the incremental pore fluid pressures. Similarly.52) can be treated in a similar manner to obtain the equation: \Apf AsvdVol=f.366 / Finite element analysis in geotechnical engineering: Theory where: = I Npiit *]%CQsI0+Aplfi sin/60 i=\ i\ 1=0 10 and Ap'fl and Aplfi are respectively the 7th cosine and sine coefficients of incremental pore fluid pressures for the fh node. The cosine harmonic coefficients of incremental pore fluid pressure are only associated with the parallel loads. hence they are called the parallel components of incremental pore fluid pressure.54) r}T[JBl.55) The first integral in Equation (12.56) The second integral in Equation (12. ' =1 f £ (12.']T{77l} Npj Ap'g r&area where [JL^]7 = 2TT J r [51/] T {i/l} NpJ darea ifl=0. the sine coefficients are the orthogonal components of incremental pore fluid pressure: A^J" = Ap1} and A^f = Ap1** (12.

r darea —IXIL ^ r and (Ap%)T[El°]T([ku]{iGl} +[kn]{iG2}) r darea ^{n.} (1259) darea (12.58) can be written as: T m Vpf t where: i=lJ=l (12. Accordingly. Equation (12. 0 ] 7 ([A. with the gravity term only affecting the 0th harmonic.]rdan«i 'Ji r'liEi1^ L-[*2il o J W ] r darea .Fourier series aided finite element method / 367 The [A:] matrix is split into four parts in accordance with the two parts of the [E\ matrix and the d6 integral is carried out.58) separates the components that are symmetrically uncoupled from those that are coupled. Inspection of these terms reveals that the condition for symmetric uncoupling is that the [kl2] and [k2]] components of the [A:] matrix are zero.58) where {«.. The hydraulic gradient has been divided into to two parts.} = 2n J [^l. consisting of the pore fluid pressure and gravity components. This is correct if gravity acts in a fixed direction in the r-z plane (i. If the direction of gravity acts out of this plane and is dependent on r and 6. the equations became more complex.60) [ < ] = 2JLj[£i?]T[*11][Ei. Equation (12.]{/G1} + [h^iki)) r darea.e. The dashed line in Equation (12. Both equations are harmonically uncoupled.57) can be written in two parts: 1 [k]Vpf dVol =—jSZ Yf i=lj=l l . z direction).

with the orthogonal and parallel equations having the same stiffness contributions. The same equation can be used for the orthogonal Ith harmonic. the resulting equations are also symmetrically uncoupled. i.64) .52) and write them in matrix form.61) where: pf0 is pf at time t = t0. Apfis the change in pf over At. Accordingly.368 / Finite element analysis in geotechnical engineering: Theory A simple time stepping approach is used to carry out the integral from time t0 to t = to+At.51)and (12. In this case [k] can be written as a Fourier series: [A] = [ '] cos0+[k] ']sin2<9+ . Thus for any harmonic order we can combine Equations (12. see Chapter 10: (12.63) The material permeability may not be a constant and could vary with stress/strain level. However. = 0 otherwise.e. + (12. For the parallel 7th harmonic we get: (12.52) are harmonically uncoupled..62) -j3At[<PG] ]" where «/i c )) = nG if / =0 . if either the [D] matrix or the [k] matrix do not satisfy the symmetry criterion. If the criterion for symmetric uncoupling is satisfied by both the [D] and the [A:] matrices. both orthogonal and parallel coefficients have to be solved simultaneously: 0 - 0 [K'GY [L'G] (12. [k] could vary in the circumferential direction. p/v=pfQ+pApf.. all the components of Equations (12. For the simple constant [D] and constant [k] case.51) and (12. p is the parameter that defines the average pore water pressure over the time step.

= Vi otherwise./ ] + /3[*1V/] .65) V/=o k=\ where: a!*.l<\. Similarly.59) now involves the integral of the product of two Fourier series.+'] ±«l:*iV']-y9[**2+'] y. + i s ."']+>8[*. • r darea 1 +/ ±a[**.*. [k1] and [kl] are matrices containing the 0th. «[*n"']-A*n+'] ±«[ 4' •./> 0. is . =Vi otherwise. P = 1 if k = I = 0.if^.r > r darea T«[4']-A4'] « :4']-A4'i. The solution is harmonically coupled and is similar to that for a variable [D] matrix: \V{Apf)J[k]VpfdVol = 1=0 1=\ k=\ (12.66) ./ > 0.Fourier series aided finite element method / 369 where [k°].«[Af2-'] + / 9[4 ] - fl[ *" ^ ' 1+ t 1 + ±a |_/r99 J — p\k^^ \ +a[Af.'] ^[4V^2r] ]+#*£'] [in*] r darea « [ 4VA4']. /th order cosine and Ith order sine harmonic coefficients of the components of the [A:] matrix respectively. thus the gravity term affects all the harmonics: r darea (12./ < 1. ±is + i f £ .*. the solution for Equation (12. is + if A .-'] + yS[Af1 '] «| r darea In the above equations: a = 1 if k = I.52) now involves the integral of the product of three Fourier series. The solution to Equation (12.if*.

4 Implementation of the CFSAFEM MA A Introduction From the formulation provided in the previous sections it is clear that the CFSAFEM involves many extensions and enhancements of the conventional finite element theory presented in Chapters 2. In both cases appropriate correction to the right hand side of the finite element equations will be necessary if [A:] is variable. due to . an equation with a form similar to that of Equation (12. while the conventional finite element formulation is in terms of real values. the two additional components of stress and strain and the use of a number of harmonic coefficients in a CFSAFEM analysis. elasto-plastic). In particular. However. the boundary conditions and the output from an analysis. However. in which the harmonically coupled terms are ignored. the p^ terms now have to be multiplied with all the harmonically coupled [0G] matrices.e. There is a large increase in data storage requirements compared to a conventional 2D finite element analysis. 10 and 11. As discussed previously regarding the nonlinear material behaviour (i. The general criterion for symmetric uncoupling in this case is that the [ku] and [k22] parts of the [£] matrix are symmetric functions of 6.63) can be written to relate any 7th harmonic right hand side with any kthharmonic left hand side: /=0 7=0 where ((([Lj]))) = [Lj] if / = k . = 0 otherwise.370 / Finite element analysis in geotechnical engineering: Theory The [LG] terms in Equation (12. CFSAFEM analysis of coupled problems can be performed either using the constant [k] formulation or a compromised variable [k] formulation. Since the equations are now harmonically coupled. 12. the CFSAFEM formulation is expressed in terms of Fourier series coefficients. Inclusion of the CFSAFEM into an existing finite element program involves a considerable effort. This clearly has implications for the computer code.63) are unchanged since they are not affected by [k] or [D]. and the [k]2] and [k2l] parts are asymmetric. This is due to the additional coordinate direction.

4. Thus a general approach has to be devised for expressing the distribution of variables in the 6 direction as a Fourier series. .3. since an assumption has to be made regarding the value of x between any two specified values.2 Evaluating Fourier series harmonic coefficients All the variables used in a CFSAFEM analysis have to be expressed as Fourier series.e. =-J/(i9)sin/<9d6> Often there is no explicit expression for x.4. If there is an explicit expression for x. x =..Fourier series aided finite element method / 371 the harmonic uncoupling this storage requirement is likely to be considerably smaller than that required for a comparable conventional 3D analysis.... Two methods are suggested: (i) the stepwise linear method and (ii) the fitted method.4. + A cos/0+A sin/0 + . Significant modifications must also be made to the nonlinear solution algorithm to enable the correct adjustment to be made to the right hand side of the finite element equations. where n is the number of known values of x. It is also used to interpret complex boundary conditions and to formulate the partial nonlinear stiffness matrix. The material presented is based on the Authors' experiences with implementing the CFSAFEM into the computer code ICFEP. .. . e. instead values of x at specific 6 values are known. x3 at 03.g. xf at #. then harmonic coefficients can be evaluated from the equations: 2% _J/ X1 = — f / ( # ) coslO&O X .. x} at 6{.3.. This is a key component of the MNR solution strategy for nonlinear CFSAFEM. xn at 6n. Methods of specifying input boundary conditions also have to be enhanced to provide a user friendly interface with the finite element program. as described in Section 12. because the CFSAFEM is formulated entirely in terms of Fourier series harmonic coefficients. In the following subsections some of these topics are considered in more detail.. (12 69) r/ r/ (12 68) ./(#). The harmonic coefficients have to be evaluated such that x can be expressed as: . i. There is no unique solution to this type of problem. see Section 12. Efficient data management routines are required to store data and to convert it between harmonic coefficient and real values. x2 at 62. 12.. since it is used to evaluate the right hand side corrective loads. Consider a variable x which is a function of 6.

The integration is performed in Appendix XII. Using the known values of x they assume a stepwise linear distribution with 6.cos£6j. This is impractical in any numerical algorithm and invariably a finite number of terms is used.70) Equations (12. even though the resulting Fourier series does not fit the stepwise linear distribution exactly.4. .4: Step wise linear method carrying out the integration analytically. see Figure 12. if an infinite number of harmonic terms were used in expression (12.71) where: X n+] = X \ The harmonic coefficients obtained from Equation (12.68).1 The stepwise linear method This method is based on an approach suggested by Winnicki and Zienkiewicz (1979).372 / Finite element analysis in geotechnical engineering: Theory 12. Thus the value of x at any value of 9 is: (12. Additionally. This has the advantage of both eliminating any integration errors and reducing the computer resources required.69) are then integrated numerically to obtain the harmonic coefficients and Bode's integration rule is suggested..4.6 and the solutions are presented below: £ -+1 .69) would fit any continuous function exactly. Stepwise linear distribution Fourier series representation y This method is improved upon by Figure 12.) (12. which may even be better than the stepwise distribution. Usually a truncated Fourier series is used where the higher order terms are ignored and sufficient terms are considered such that the error associated with ignoring these higher order terms is small.2. (*. There is a trade off between the error associated with this numerical integration and the amount of computer resources required to evaluate it. it will still give a good representation of x.

2 The fitted method In this method it is assumed that x can be represented by a Fourier series that passes through all known values of x. where n is the number of unknown values specified. The harmonic coefficients for the unique solution can be obtained by substituting each known value of x and 6 into Equation (12. can now be solved for by inverting [//] and premultiplying x by [H]~l. n. sin<9. The number of harmonic terms used in the Fourier series representation is an important Fourier series representation parameter.68). Solutions have been derived for the special case of having known x values at equispaced 6 values. If there are less harmonic terms._.5: Fitted method number of possible solutions exist..73) where x is the vector of known values x>. see Figure 12. s'mOn cosLOn sinL6)n X2 X w-1 2 where L is the order of the truncated Fourier series used (equal to Vi(n-\)) and can be rewritten as: x = [H]X (12..72) 1 cos<9/. Thus a number of AT corresponding to a number of JC can be obtained very easily by premultiplying the various x by the same [//]"'. This condition is achieved by using a truncated Fourier series. and can be expressed in matrix form as: . X is the vector of unknown harmonic coefficients. is assumed to be odd such that the resulting truncated Fourier series has the same . sin<92 ._.5. a solution does not exist in general. If the 6 values at which x is known are unchanged. A unique solution exists if the number of harmonic coefficients is equal to the number of known values. then [//] is unchanged.2. see Appendix XII. a Figure 12.4.7. [H] is the harmonic transformation matrix that is a function of the 6 values at which x is known..Fourier series aided finite element method / 373 12. The number of known values. If there are more harmonic terms than known values of x. cosL02 sin x sinLO] sinLO2 s'mLB3 x° X7 (12.. cos<92 1 cos . X. This yields n equations with n unknowns. The harmonic coefficients.

For any imposed boundary condition a trial solution is obtained by solving the system equations associated with the current global stiffness matrix.3 The modified Newton-Raphson solution strategy / 2. or Fi9ure 126: Location of x values for the location of the x values. using the material constitutive laws. This procedure is very similar to the MNR procedure for conventional nonlinear finite elements. JC1? is at 6 = 0°. as described earlier in this chapter. These laws are expressed in terms of real values rather than harmonic coefficients.3. xh will occur at 6 = (i-l)a. In the equispaced fitted method remainder of this chapter these methods will simply be referred to as the fitted and stepwise approach. It is acknowledged that there is possibly an error associated with this trial solution and a corrective right hand side load is evaluated. since this approach places no limitation on either the number of harmonic terms used. If these criteria are not satisfied. the robust stepwise linear approach is recommended. The process is repeated until the corrective loads are small. only applicable for the special case when the x values are equispaced and the number of harmonic terms (2Z+1) is equal to n. It is. However.374 / Finite element analysis in geotechnical engineering: Theory number of sine and cosine harmonic terms. see Figure 12.6. This corrective load is then used to obtain the next trial solution. Thus a procedure has to be developed which enables a harmonic correction load to be evaluated from the harmonic trial solution. the subsequent known x values. The harmonic coefficients can then be evaluated using equations: Xk =- (12. The material constitutive laws are used in the process of calculating these corrective loads.74) xk=±£ This equispaced fitted approach is the most efficient method for 2TT obtaining harmonic coefficients. 12. the CFSAFEM formulation is in terms of Fourier series coefficients.4. Thus the corrective right hand side loads have to be expressed as Fourier series and are evaluated from the trial solutions which are also expressed as Fourier series. however.1 In troduc tion The MNR solution strategy for nonlinear CFSAFEM is an iterative procedure. see Chapter 9. If the first x value. .4.

Thus it is proposed that the number of slices should be the same as the number of harmonic coefficients used to represent forces and displacements in a CFSAFEM analysis.2. the stress changes associated with the solution displacements. obtained from the imposed boundary conditions. which gives the appropriate stresses on each slice. There is no definitive method of obtaining the Fourier series representation of stress from the stresses at the sampling points. there are no guidelines for the number and location of the slices in the 6 direction. Harmonic coefficients of stress of a higher order would not influence the CFSAFEM analysis. accordingly. with the coefficients of nodal load consistent with the trial solution displacements. Conversely. a Fourier series representation of stress could be obtained. but requires the stresses to be expressed as a Fourier series in the 9 direction. Theoretically. The d8 integral is carried out analytically. the sampling points are located at Gaussian integration points in the r-z plane. Similarly.2 High t hand side correc tion The corrective right hand side load is expressed as Fourier series coefficients of nodal load. increasing the number of slices means that the Fourier series would not be able to represent the stresses on all the slices exactly and the benefit of these additional slices cannot be fully realised.3. Their locations within each slice is determined by the Gaussian integration order used and the finite element mesh.6. This is performed by substituting the harmonic solution displacement into Equation (12.4.2. see Section 9. The stresses at the sampling points are used to evaluate this integral. If there were the same number of slices as harmonic coefficients of stress in an analysis. The number of slices used in an analysis should be influenced by the order of the Fourier series required to represent stresses. as shown in Appendix XII.Fourier series aided finite element method / 375 1 2. A volume integral has to be carried out to obtain the harmonic coefficients of nodal load consistent with the stress state in the problem domain. To perform the d8 integral it is necessary to know the 8 coordinates of the sampling points. the solution of which is presented in Appendix XII. .3.8) to calculate the real incremental strains at the sampling points. They are evaluated by comparing the coefficients of nodal load. where 'area' is the area on the r-z plane. It is therefore convenient to situate the sampling points on a number of constant 6 planes (slices). The Fourier series used to represent stresses in a CFSAFEM analysis should have the same order as the Fourier series used to represent forces and displacements. for obtaining harmonics can be used.2. It is also suggested that the slices are equispaced around the circumference. in which case the efficient fitted method. The darea integral is carried out using a Gaussian integration technique and. The integral is divided into two parts.4. a dO integral and a darea integral. The latter loads are obtained by calculating. see Section 12. A rational approach is suggested below.3. at a number of sampling points within the problem domain. increasing the number of harmonics would not increase the Fourier series accuracy since the stresses on the slices could already be represented exactly. A stress point algorithm is used to integrate the material constitutive laws along these strain paths to obtain the incremental stresses.

1. see Figure 12. in the nonlinear solution strategy ensure that the state variables are consistent with the prescribed strain path. Consequently. in practice. and to be consistent with the storage of nodal variables. however the calculation of the harmonic coefficients of nodal force. see Chapter 9 and Appendix IX. it seems sensible that these values are stored as harmonic coefficients and converted to real values when required for output purposes. nodal displacement. The only decision then is in which form (real values or harmonic coefficients) should each variable be stored. see Figure 12. it is both very inefficient and cumbersome to store data in two different forms. The magnitude of this transformation error reduces as the number of harmonics used increases and. e. For these reasons. it is critical that the exact real values are evaluated. The decision is more difficult for the state variables stored at integration points. The stress point algorithm. As the nodal variables in a CFSAFEM analyses are expressed purely in terms of Fourier series harmonic coefficients.4. as opposed to a single quantity. for each nodal variable. e. nodal displacement or force. strains and hardening parameters. consistent with the stresses at the sampling points. stress or strain component) is needed either in harmonic form or as real value. so that their values at any 6 value can be exactly calculated. see Section 12. The stress point algorithm requires these variables to be expressed as real values. When the stress point algorithms next require these variables. Likewise. and in practice the projecting back subalgorithm. efficient data management routines are required to store the data in one form and to convert between harmonic coefficients and real values and vice versa.g. These real values of the state variables are then expressed and stored as harmonic coefficients. . for each variable at an integration point. it is necessary to store a number of quantities. it is necessary to store a number of harmonic coefficients. since even a small error could result in an illegal stress state. Firstly. A situation where the transformation error would be important is in the storage of the soil state variables at sampling points on the predefined slices. e. when they are used to evaluate the real values that they were derived from.g. ready for use by the CFSAFEM algorithms and the data output algorithms.g. either in the form of harmonic coefficients. Coefficients obtained using the stepwise method do not exhibit this property and there is a slight error.376 / Finite element analysis in geotechnical engineering: Theory MA A Data storage The data storage required for CFSAFEM analysis differs from that of conventional finite element analysis in two important aspects. e. should be small. The data output algorithms are also likely to require that these variables are expressed as Fourier series.5. stress component. stresses. or as real values one for each slice. However.g.g.2. 8. This interchangeability between the harmonic coefficients and real values dispenses of the need to store both harmonic coefficients and real values of any variable. An important property of any coefficients obtained using the fitted method is that real values calculated from these coefficients are exactly equal to the real values the coefficients were derived from. it seems appropriate to store the harmonic coefficients. requires that the stresses are expressed as Fourier series. At different stages during the analysis each variable (e.4.

3. The computer program then uses the stepwise method (see Section 12. Consequently.4. for each integration point the real [D] and [K] matrices are calculated at the associated circumferential series of sampling points.1) to obtain the harmonic coefficients. However. It is for this reason that the fitted method is recommended in preference to the stepwise method when calculating the load vector for the right hand side correction to the finite element equations. Three options for specifying boundary conditions are evident: The user determines the Fourier series that is equivalent (or a good approximation) to the real boundary condition.5 Boundary conditions As the CFSAFEM is formulated in terms of Fourier series coefficients.4 and 12. 12.g.4.4.3. Examples of their use are given in Volume 2 of this book. The user inputs two sets of numbers. The first set contains values of the boundary condition at a series of 9 values. The number of values must be equal to the number of Fourier coefficients required.2. and inputs the harmonic coefficients directly.4. for most practical boundary value problems the boundary conditions are likely to be in terms of real values (e. a displacement or force in a certain direction). requires that the constitutive matrix [D] and permeability matrix [K\ are expressed as a Fourier series. before any analysis can be undertaken. The computer program can then use the fitted method (see Section 12. as described in Sections 12. The user inputs a set of numbers representing the value of the boundary condition at equispaced values of 6. To achieve this.Fourier series aided finite element method / 377 stresses exceeding yield. These harmonic matrices are required for every integration point located in the r-z plane. The partial nonlinear no symmetry CFSAFEM formulation results in a nonsymmetric stiffness matrix when the stiffness contribution for a parallel symmetry load on an orthogonal symmetry displacement is not the same as that for the corresponding orthogonal symmetry load on the corresponding parallel symmetry . which could result in large errors. The harmonic coefficients for these matrices can then be obtained using the fitted method.7. The second set contains the magnitudes of these 6 values. i. all boundary conditions must be specified in this manner. these real values must be converted to an equivalent Fourier series representation.2) to obtain the required harmonic coefficients. - In the Authors' experience all three of the above options should be available to the user of a computer program. In this case the stress point and other associated algorithms would not be valid and some arbitrary form of correction is required.6 Stiffness matrices The partial nonlinear formulation. on each slice. 12.e.2.

As this situation is generally applicable. The criterion for this second requirement concerns the nature of the constitutive matrix [D] and the permeability matrix [A:]. as noted in Section 12. this approach of considering only parallel or orthogonal terms is theoretically valid only if two requirements are met.378 / Finite element analysis in geotechnical engineering: Theory displacement.3. The first requirement is that the boundary conditions satisfy the appropriate symmetries. such that the imposed boundary conditions consist of purely parallel or orthogonal symmetry terms. many boundary value problems have a symmetry about the 0 = 0° direction.4 and 12. for orthogonal symmetry the components of the boundary conditions in the r and z direction have an asymmetric variation and those in 6 direction have a symmetric variation with 6. or orthogonal symmetry (e.7.3. In all other situations. whereas the components in the 6 direction have an asymmetric variation. An example highlighting this problem is presented in the next section. However.4. 1 2. by ignoring the harmonics associated with the zero coefficients. In the parallel symmetry case all components of the boundary conditions in the r and z coordinate directions have a symmetric variation with 8. 12. For example. Consequently.4.g. see Sections 12. for all the r and z components of the boundary conditions the sine terms in their Fourier series have zero coefficients. While such approaches restrict the type of boundary value problem that can be analysed.1 Introduction As noted in Section 12. Winnicki and Zienkiewicz (1979)). even if the boundary conditions satisfy the symmetry requirements. the symmetry of the global stiffness matrix depends on the nature of the constitutive matrix [D] and permeability matrix [A]. using the partial nonlinear CFSAFEM. and for the 6 components the zeroth and cosine terms have zero coefficients. must be performed.2.3.2. the global stiffness matrix must be inverted using a non-symmetric solver. Consequently.3. .37. the global stiffness matrix is non-symmetric. If the second condition is not satisfied. and for sine terms associated with the components in the 6 direction. In such a situation a full non-symmetric analysis. Conversely. This saving arises due to the fact that only half of the Fourier series terms are considered. 12. Griffiths and Lane (1990)). This results in zero values for the coefficients of the zeroth and cosine terms associated with the r and z components of the boundary conditions. 7.g.2. almost half of the harmonic coefficients are zero.3.7 Simplifications due to symmetrical boundary conditions 12. it is then incorrect to perform an analysis with only parallel or orthogonal symmetry terms. accounting for all terms in the Fourier series. they lead to a large saving in computer resources required for an analysis and significantly simplify the solution algorithms. and the second requirement is that the system equations are symmetrically uncoupled. for either a parallel or orthogonal symmetry problem. if a non-associated elasto-plastic constitutive model is employed.3 A and 12. The majority of previous implementations of the CFSAFEM have capitalised on the above consequences and have been coded to deal only with either parallel symmetry (e.

Fourier series aided finite element method / 379 Initial stresses can also present potential problems for parallel and orthogonal symmetry analysis. and Ac^) yield parallel loads and zeroth harmonic components of {Ac2} (i. The flow terms resulting from this regime counterbalance the gravity flow terms such that there are no net symmetrical right hand side flow terms. a parallel analysis is valid if the stress changes do not yield an orthogonal load. {iG}.) yield orthogonal loads.e. Consider the situation in which the initial stresses in the soil do not vary in the 6 direction.. orthogonal and no symmetry. due to the uncertainties involved with satisfying the requirements concerning the nature of the constitutive matrix [/>]. 0° < 6 < 360°) must be used. . from Appendix XII. while in a no symmetry analysis the full geometry (i. At the beginning of an analysis.e. A<7Z. If the symmetry criteria cannot be satisfied.3. gives rise to a zeroth harmonic flow term on the right hand side of the finite element equations. This conflict can be resolved by specifying a hydrostatic initial pore fluid pressure regime. AT. AT. If these conditions are not satisfied. In a linear consolidation CFSAFEM analysis the gravity vector. Similarly an orthogonal analysis is valid if there are no stress changes giving parallel loads. it is often useful to compare no symmetry and the relevant parallel or orthogonal symmetry analysis.2. It is the Authors' experience that all three options. A no symmetry analysis is also required if on first loading from the initial stress state the constitutive matrix [D] depends on the initial state of stress. a no symmetry analysis should be used. A<7. This questions the validity of {(xl} initial stresses in an orthogonal analysis and {al} initial stresses in a parallel analysis. Also. which is likely to be the case for analysing nonlinear problems. Here all the stress components can be represented by a Fourier series when all the harmonic coefficients.g.3. This problem is partly overcome if an incremental form of the CFSAFEM is implemented. zeroth harmonic components of {A<rl} (i. Although a no symmetry analysis is considerably more expensive in terms of computer resources than either a parallel or orthogonal analysis. there is a conflict between the gravity zero harmonic terms and the orthogonal symmetry criterion for an asymmetrical flow term.. see the example given below in Section 12. the initial stresses are assumed to be in equilibrium with themselves and with the initial boundary stresses. The incremental form of CFSAFEM then evaluates the stress changes.# and Arr6. are useful and should be available in a general purpose finite element code. irrespective of the actual values of stress. However. except the zeroth harmonic...e. many problems have to be analysed in this way because they do not satisfy all the requirements for either a parallel or orthogonal analysis. are zero. parallel. However. a no symmetry analysis should be undertaken. see Section 12. a linear or nonlinear consolidation CFSAFEM requires that the initial pore fluid pressures and {ia} are consistent with the symmetry of the analysis. e. It should be noted that when analysing problems with parallel or orthogonal symmetry only half of the geometry (i.e.. to confirm that the requirements are satisfied.4.7.7. 0° < 6 < 180°) needs to be considered. More generally. Thus. which should be consistent with the applied nodal loads. This is consistent with the parallel symmetry criterion for a symmetrical flow term.

The results of both the CFSAFEM analyses (i. Since the inner and outer pressures remain unchanged. This problem involves the analysis of a hollow cylinder soil sample subjected to a torsional displacement. No restriction is imposed on the radial movements. Gens and Potts (1984) analysed this problem using a special quasi axisymmetric finite element formulation. stresses and Figure 12. with only the zeroth harmonic terms.7.1. However. undrained analyses are performed by specifying an equivalent bulk modulus for the pore fluid. To ensure strain compatibility. the axial strain is kept to zero and it is Finite element mesh assumed that there are no end Hollow cylinder sample effects.7: Hollow cylinder sample strains are independent of subjected to torsion coordinates z and 9 and it is therefore possible to analyse a section of the sample as shown in Figure 12. Ke. see Section 3. As a consequence of this latter assumption. while maintaining the inner and outer pressures equal and constant.4. In these analyses the modified Cam clay model is used to represent the soil and the material parameters and initial stress conditions are given in Table 12. Also shown in this figure is the finite element mesh used to perform this analysis. It represents a section of the hollow cylinder wall. orthogonal and no symmetry) are identical and are in agreement with those of . Analyses are therefore performed using both an orthogonal symmetry and a no symmetry CFSAFEM formulation and the results are compared with those of Gens and Potts (1984). two examples are presented.380 / Finite element analysis in geotechnical engineering: Theory 12. to be 100 times the bulk modulus of the soil skeleton. but the vertical movements along AB and CD are set to zero. with an inner radius of 100 mm and an outer radius of 125 mm. In addition.7. and to emphasise the importance of having the option for a no symmetry analysis. whereas the same displacement component is held to zero along AB. To simulate the test conditions a circumferential displacement is applied along the boundary CD. The wall is divided into ten eight noded isoparametric elements of equal radial size. Initially. the boundary conditions described above can be presented using orthogonal symmetry. the applied displacements along CD have to be proportional to the value of the r coordinate.2 Examples of problems associated with parallel and orthogonal analysis To illustrate some of the problems associated with implementing a CFSAFEM formulation for analysing either purely parallel or orthogonal symmetry problems.e.4. a zero normal pressure change is applied during the analysis to boundaries AD and BC. The first example considers the hollow cylinder problem described by Gens and Potts (1984).

The results tend towards the undrained analysis results. Mj Isotropic initial stress Elastic shear modulus. These results show a much softer response and a higher ultimate load to those from the undrained analyses presented in Figure 12. This can be seen in Figure 12.0077 0. these are defined to be zero in an orthogonal CFSAFEM formulation.8.693 200kPa 18675kPa 0.0 1.8: Undrained hollow cylinder test + Quasi-axisymmetric Orthogonal No symmetry 1 2 Rotation (10° rad) Figure 12.066 0. X Slope of swelling line in v-lnp' space. even though the boundary conditions can be specified using orthogonal symmetry. The reason for the discrepancy is that in a drained problem the zeroth harmonic coefficients of radial displacement are non-zero.e. Ke . The analyses are repeated assuming drained soil conditions. all parameters and boundary conditions the same.788 0. which are similar to those from the undrained analyses.Fourier series aided finite element method / 381 Gens and Potts (1984).e. Figure 12. K Slope of critical state line in J-p' space. twist).9. The result of the two CFSAFEM analyses are compared with the results from the quasi axi-symmetric analyses in Figure 12.8.2 Rotation (10"3rad) Figure 12. which shows the variation of the moment. which. due to the incompressible material behaviour (i. again in terms of moment versus rotation.e.1: Material properties for hollow cylinder problem Overconsolidation ratio Specific volume at unit pressure on virgin consolidation line. with angular rotation (i. the orthogonal CFSAFEM analysis gives very different results. i. G It can be seen that the quasi axisymmetric and no symmetry CFSAFEM results are identical. Thus. However. except that the effective bulk modulus of the pore fluid. 1. is set to zero.9: Drained hollow cylinder test . the correct drained solution cannot be obtained.1 0. However. vx Slope of virgin consolidation line in v-\np' space. applied to the top of the sample.

Pisa* view hence the problem can be analysed using both the CFSAFEM and a plane strain analysis. e. The second example considers a rigid horizontal disk pushed laterally through a soil mass with a circular boundary.10. with a uniform displacement in the 0=0° direction imposed on the rigid disk. For the I1 1 I I _L present example ten harmonics are used.11: Boundary conditions and meshes for pile section symmetry. Schematic view of horizontally loaded pile section conditions and typical meshes used for both types of analyses are shown in Figure 12. This is an example of a problem where a no symmetry analysis is required due to the material properties. A displacement S is specified by prescribing the nodal displacements at the disk-soil interface to be: . For the present example the problem is displacement controlled. However. the boundary for CFSAFEM pile section analysis conditions imposed have parallel symmetry. see Section 12.g. Only the first Fourier harmonic coefficient is required to represent the boundary a) Boundary conditions and mesh used for plane strain pile section analysis conditions for the CFSAFEM analysis.382 / Finite element analysis in geotechnical engineering: Theory high Ke value). The boundary Figure 12. even though the boundary conditions can be specified using orthogonal symmetry. Interface elements are located at the disk-soil boundary so that various interface conditions. have no radial displacements and under predict the correct drained ultimate load by over 40%. rough with breakway.3 A. © © = -8 sine u = 6 cos0 i—1> have to be represented using more r. The disk diameter.J). smooth. D. if it is in the 6=K/2 direction they have orthogonal Figure 12. u harmonics. This geometry is representative of a horizontal section across a laterally loaded pile. If the direction of loading is in b) Boundary conditions and mesh used the 0=0° direction.11. rough. can be investigated. is 2m and the boundary diameter is 40m.10. smooth with breakway. and if it is in any other analyses direction they have no symmetry.v analysis the solution displacements t . in a nonlinear z. as shown in Figure 12. It is assumed that there are no vertical strains.

12.07 kN. The shear strength of the interface elements is also assigned a value of 75.94SUD (Randolph and Houlsby (1984)). but there is a different initial stress regime: a. To keep the undrained shear strength and thus the limit load the same as in case 1..e. The CFSAFEM analyses were carried out using the parallel symmetry and no symmetry options and the load displacement curves are shown in Figure 12. The reaction force on the disk due to the imposed displacement can be evaluated. u= S cosO .3 and has 1500 an initial isotropic stress state with Gf.'= av' = o0' = 138.12 kPa. easel 200 kPa. wherein only the parallel symmetry terms in the formulation are implemented. To illustrate the problems associated with the past implementations of the CFSAFEM. see Section 7. For these conditions. Ke.13.'=a2'=a0'=200 kPa. the undrained shear strength for the soil is 75.12: Load-displacement and an initial isotropic stress state of curves for pile section. The plane strain analyses required the initial stresses to be expressed in the global xG and yG directions. the pre-consolidation pressure. the soil is modelled as 2000 modified Cam clay.4. The no symmetry option agreed with the plane strain analysis. Circumferential displacement.Fourier series aided finite element method / 383 Radial displacement.4.3 Figure 12. see Appendix XII. was performed. but .12 kPa. po\ was set to 210 kPa. using the soil properties given in Table 9. The net reaction force in the 0=0° direction can be obtained by adding the first cosine harmonic coefficients of radial nodal reactions and subtracting the first sine harmonic coefficients of circumferential nodal reactions. a further analysis. giving a value of 1794. thus at any point the stresses defined above had to be rotated in accordance with their location. as shown in Figure 12. since an associated flow rule is used. It is assumed to behave undrained during loading and this is simulated by specifying a high bulk modulus of the pore fluid. For case 1. co = -8 sin#. The soil fails under plane strain conditions and. see Appendix VII. the soil properties given in Table 9. The results show that both parallel symmetry and no symmetry CFSAFEM analyses are identical and are within half a percent of the plane strain analysis. This also results in the soil adjacent to the pile being normally consolidated. where r is the radial coordinate and an anticlockwise positive convention for 6 is used. as for the first example. The limit load achieved is within one percent of the analytical solution. 20 40 60 Displacement (mm) the Lode's angle at failure is 0°. In case 2 the soil properties are the same as case 1.0/r kPa. The analytical limit load for this problem is 1 \.12. case 2.22 kPa and xrB = -69. i.

the [Dep] matrix for case 2 of the normally consolidated material adjacent to the pile. except that the shear stress is now applied in the opposite direction. Consider. for the problem considered. for case 3 the [Dep] matrix for the material adjacent to the pile is: . i. however.13: Load-displacement the parallel symmetry analysis.e.13. Thus the criterion that the [Dl2ep] and [D2lep] terms are asymmetrical functions of 6 is not satisfied and a no symmetry CFSAFEM analysis is required. This is an example of a problem where a no symmetry analysis is required due to the material properties. curves for pile section.1 for the definition of [Dep]. and [D]2P] and [D2]ep] terms that are asymmetric functions of 0. see Section 12. The load-displacement curve for Displacement (mm) the no symmetry analysis is the same as for case 2.0/r 80 20 40 60 kPa. There is still an error in Figure 12. A no symmetry analysis should be undertaken if there is any doubt about the validity of the symmetry criterion. at the beginning of the analysis: 48990 11640 11640 11640 48990 11640 11640 11640 48990 0 0 0 0 0 -5402 -5402 -5402 1770 0 0 ep 0 0 0 0 18680 -5402 -5402 -5402 -5402 0 0 0 0 0 18680 Du\Dl2 ~D~\b~2 0 This [D ] matrix is constant in the 6 direction and hence is equal to the zeroth harmonic [Depf.384 / Finite element analysis in geotechnical engineering: Theory there was a significant error in the parallel symmetry analysis. The limit loads from the no symmetry and Limit load Parallel symmetry: case 2 plane strain analyses agreed with the No symmetry: case 2 & 3 analytical solution. even though the boundary conditions could be specified using parallel symmetry. The errors in both case 2 and case 3 parallel symmetry analyses can be attributed to the symmetrically coupled nature of the soil. see Figure 12.3. These results emphasise the importance of the implementation of the no symmetry CFSAFEM formulation. discussed in Section 123 A. the load deflection curve is case 2 and 3 now softer than for the no symmetry analyses and the same limit load is reached. xrB =69. The symmetry criterion is that the material elasto-plastic [Dep] matrix has [Duep] and [D22ep] terms that are symmetric functions of 6. Similarly. for example. Parallel symmetry: case 3 Plane strain Case 3 is similar to case 2. which was able to analyse both case 2 and case 3 accurately. which 1500 gave a stiffer response initially and reached a higher limit load.

5. 0). for each 2D axi-symmetric node. -0) would be the normally consolidated stress state described in case 3. The discrete Fourier series aided finite element method 12. during loading of case 1. Note that though the [Dl2p] and [D2lep] terms may be non-zero at any particular 6 coordinate. 12.5 . However. In this instance. If the loading in case 1 causes a stress change at a point (r. the criteria for symmetrical uncoupling are still valid.1 Introduction The discrete FSAFEM is similar to the continuous FSAFEM and reduces a large 3D global stiffness matrix into a series of smaller uncoupled submatrices. such that its stress state was equal to the normally consolidated stress state described in case 2. the zeroth and cosine harmonic coefficients of [Dx epr '] and [D2r] are zero. the continuous method considers a 2D axi-symmetric mesh and employs a continuous Fourier series to represent the variation of force and displacement in the circumferential direction. the symmetry criterion is still satisfied. The [Dep] matrix at the beginning of the case 1 analysis is: 29260 -8091 -8091 0 0 0 0 0 -8091 29260 -8091 0 0 -8091 29260 0 0 -8091 [Dep] = [Dep]° = 0 0 0 0 18670 0 0 18670 0 0 0 0 0 18670 0 0 0 0 Thus before loading the criterion for symmetrical uncoupling is satisfied and a parallel symmetry analysis can be used. the discrete method considers a full 3D finite element mesh and employs a discrete Fourier series to represent the nodal values of force and displacement for the succession of nodes on each circumference. corresponding to each 2D axisymmetric node. In contrast. The [Dep] matrices for case 2 and case 3 also suggest that.Fourier series aided finite element method / 385 48990 11640 11640 0 5402 0 11640 48990 11640 0 5402 0 11640 0 |5402 0 11640 0 i5402 0 48990 0 J5402 0 0 18680J5402 0 5402 0 !l770 0 i 0 0 ! o i8680 Thus the criterion that the [D]2P] and [D2Xep] terms are asymmetrical functions of 6 is also not satisfied and a no symmetry CFSAFEM analysis is again required. and the symmetry criterion is satisfied. the stress state at point (r. since the [Duep] and [D22ep] terms would be symmetric functions of 6 and the [Dl2ep] and [D2lcp] terms would be asymmetric functions of 6.

The method can be expanded to consider twenty node linear strain brick elements. This is similar to a finite element mesh for a .14. and such an extension is presented in Lai and Booker (1991). in order to simplify the description. V H z r a typical element a) Elevation view of a typical vertical section 71-2 a typical wedge of elements plane i b) Plan view of a typical horizontal section Figure 12. Subsequently.14: Finite element discretisation for discrete FSAFEM The problem domain is discretised using a mesh of hexahedral finite elements consisting of n evenly spaced wedges around the circular boundary in the 6 direction.2 Description of the discrete FSAFEM method The discrete FSAFEM described in this section is for an analysis consisting of eight node constant strain brick elements.5. as shown in Figure 12.386 / Finite element analysis in geotechnical engineering: Theory The first application of the discrete FSAFEM approach in geotechnical engineering was by Moore and Booker (1982) to develop a boundary element for use in deep tunnel problems. The usual practice of using complex numbers in the formulation is avoided. Lai and Booker (1991) used the method to study the behaviour of laterally loaded caissons and Runesson and Booker (1982 and 1983) used it in the study of the effects of consolidation and soil layering on soil behaviour. 12.

75) Since the material properties are assumed not to vary in the 6 direction.} and {Aw. Hence. {Ad}. i.}. Let these elements be numbered from 0 to n-1 anticlockwise and let the vertical planes be numbered from 0 to n-1. {A/?}. see Figure 12.76) can be rewritten as: Expanding Equation (12. in terms of the incremental nodal displacements of the element.} + {AM/+1}T[F]T{AW/} /•=o (12. where {Aw.-] = [A/+1].] such that: AW. The standard finite element formulation for linear elastic behaviour. to the vector of all the incremental nodal displacements.78) + {Aw.Fourier series aided finite element method / 387 full 3D analysis. [A.? {AuMf}[K. {Aw. . [Y]. Thus the internal incremental work done in the whole annular ring is: . as in Equation (12. AW.e.} + {Aw. In the discrete Fourier series aided finite element method it is assumed that the incremental nodal displacement {Aw. ={{*.78) is the equation solved directly in a full 3D analysis.}) and on plane /+1 (i. see Chapter 2.14. {Aw/+1}). Consider the shaded annular ring of n elements.Awj The stiffness matrix.e. in accordance with the two parts of the element displacements {Aw. expresses the incremental internal work. gives: AW = ''2 2 {Aw.]^^ (12. {Uk}cosika + {Uk}sinika (12.} and {Aw/+1}. the stiffness matrices for each element are identical.+1}. {Aw.}T[Z]{Aw7} Equation (12.} = {Aw//+.e. [KG]. can be partitioned into four submatrices [X\.3). done by element i. Equation (12. This form of the finite element formulation yields a single large global stiffness matrix. Thus the incremental displacements in element / are defined by the element shape functions and the incremental displacements on plane / (i.77) and using the periodicity of the axi-symmetric geometry.79) . such that the two vertical planes enclosing element / are plane / and /+1. [Y]T and [Z]. and the stiffness matrix of the element [K. which relates the vector of all the incremental nodal loads.} on plane / can be evaluated from the discrete Fourier series which represents the nodal displacement for that pseudo 2D node: {Aw.} = -jL-2._1}T[F]{AW. [K].}T[X]{A«.

80) The orthogonal properties of a discrete Fourier series are: J cosika sinila = 0 /=0 n-\ X cosika cosila .} cosila + {U.p Y {Uk) cosika + {uk) si^ta | i .81) are used to simplify Equation (12. it is only valid for integer values of/. {Uk} cos(i-\)ka +{Uk} sin(i-l)ka\ 4ni *=o J [X] | -p. Introducing expression (12.79) into Equation (12.{Z^}T[7]T{Z^}sin/a (12. . a = 2n/n.79) only describes incremental nodal displacements and cannot be used to evaluate incremental displacements at intermediate values of 0. Equation (12. thus the large global stiffness matrix has been uncoupled into n submatrices.} cosila + {U.'£ {U. = nfor I = k /=0 (12.e. i.0 for / ^ k. i.M — = Y -~Y.78) transforms the problem from determining the unknown incremental nodal displacements to that of finding the Fourier coefficients: ^w= Z| . The stiffness matrix [Kj] for any harmonic .n for / = k /=o X sinika sinila = 0 for / ^ k.} sinila {-Jn i=o ( n-\ — = ) yjn k=o J —= ^ {Uk} cosika + {Uk} sinika y/n k=o J ' K-Jn 1=0 [Z] J —= ^ {U/} cosila + {Uj} sinila {y/n /=o J (12.e.81) Equations (12.82) The internal incremental work equation has been divided into n independent expressions.} sinila | /=o (i »-i [Y] \-j=Y< {U. the equations are harmonically uncoupled.80) such that: 1=0 {t^}T[y]T{f^}cos/a + {^"}T[F]T{f7jsin/a .388 / Finite element analysis in geotechnical engineering: Theory where: {U~k} and {U~k} are respectively vectors of the kth cosine and sine harmonic coefficients of {Au} expressed as a discrete Fourier series function.4). one expression for each harmonic order. It is for this reason that the method is referred to as the 'discrete' FSAFEM. as in Equation (12.

and accordingly the harmonic coefficients of incremental displacement. and a circumferential component sub-vector: (12. a circumferential and a vertical component. This is similar to the symmetrically coupled continuous FSAFEM. {Uk} and {Uk} . a radial. Equation (12. both the sine and cosine coefficients of load and displacement have to be solved simultaneously. This is achieved by first regrouping the terms in Equation (12. can be obtained by substituting / = / in Equation (12. However. a radial and vertical component sub vector.82) as: /=0 + w.82). For an isotropic . [F]. as shown in Equation (12. for each harmonic order. /. Thus the two vectors of nodal harmonic displacements.16). and [Z] accordingly.} &} (12. ) smla smla - Lt^M Each set of wedge shaped elements defined in the mesh is geometrically symmetrical about a constant 6 plane through the middle of it. Symmetric uncoupling is possible with the discrete method for an isotropic linear elastic material.84) Expanding matrices [X\.}1 ([Y]sinla-[Y] T cosla+ [Z]) VjJ s'mla) {U.83) can be written in the following form: n-\ AW= y cosla • cosla cosla + cosla sin/a - sinla . have three components.Fourier series aided finite element method / 389 order. can each be divided into two sub-vectors.83) ([Y]smla-[Y]T s'mla) The incremental nodal displacements.

Using these terms Equation (12.86) 0 [Y22]_ 0 \Y2l] [Y] .87) can be rearranged such that: . T n] + [F n ]cos/a [A [F21]T sin/a ==\ (12-88) [F21]sin/a u] + [Yu]cosla [F 21 fsin/a [Y22]cosla [F2]]sin/a The parallel displacement terms have been uncoupled from the orthogonal displacement terms and the stiffness matrix for both set of terms is the same.87) Parallel displacement terms can be obtained by mixing cosine coefficients of radial and vertical displacement with sine coefficients of circumferential displacement.390 / Finite element analysis in geotechnical engineering: Theory elastic material this implies that the element stiffness matrix has a symmetry such that: ~[XU] 0 [X] + [Z] = 2 0 [X22] [F] + [F] T = 2 '[Yu] 0 (12. This .86) are used to simplify Equation (12.[F] T = 2 ~[Y2if 0 Equations (12.85): AW = 2% 1=0 — i (12. Similarly. orthogonal displacements are sine coefficients of radial and vertical displacement and cosine coefficients of circumferential displacement.

for example for the kth harmonic: u] + [Yu]coska (12. Parallel loads can be obtained by mixing cosine coefficients of radial and vertical loads with sine coefficients of circumferential load. The discrete method notionally considers a full three dimensional finite element mesh defined in a cylindrical coordinate system. the shape function of the pertinent element.89) [F21]sinfe [Xu]+[Yu]coska Thus the large global stiffness matrix has been harmonically and symmetrically uncoupled. Thus the force and displacement distribution in the circumferential direction is described using the standard finite element approximation.Fourier series aided finite element method / 391 is similar to the symmetrically uncoupled continuous FSAFEM stiffness matrix shown in Equation (12. This statement is misleading since it wrongly implies that the correct displacement field can be represented after a . for example for the kih harmonic: -{A<}J The principle of virtual work is then used to obtain the system equations which consists of 2n separate equations (two equations for each harmonic order). There are two vectors and one stiffness matrix for each harmonic order. orthogonal loads can be obtained. visco-plastic or MNR methods. Thus the internal incremental work equation now consists of 2/i uncoupled vectors of harmonic incremental displacement coefficients and n stiffness matrices. i.88). e. The procedure can be extended to consider nonlinear behaviour by combining the linear elastic formulation with a solution strategy which continually adjusts the right hand side of the governing finite element equations. Lai and Booker (1991) state that 'the discrete method gives an exact representation (of displacements) after a finite number of terms'. Similarly.6 Comparison between the discrete and the continuous FSAFEM A key difference between the discrete and the continuous FSAFEM method is the manner in which the variations in forces and displacements in the circumferential direction are described. The stiffness matrix for the fh harmonic order is obtained by substituting i for / in Equation (12.e.g. 12.17). The procedure outlined above is for a linear elastic analysis. The externally applied incremental loads can also be expressed as a discrete Fourier series.

and the harmonic coefficients are obtained from the sampling point stresses using the fitted method. The study compared the continuous FSAFEM. the relationship between the number of harmonics or brick elements used and the accuracy with which the resulting stress distribution can represent the chosen distribution is examined.15. Thus the Fourier series . the discrete FSAFEM with eight noded constant strain brick elements and the discrete FSAFEM with twenty noded linear strain brick elements. due to the finite element approximation.392 / Finite element analysis in geotechnical engineering: Theory finite number of Fourier series terms. Thus for both the continuous and discrete nonlinear FSAFEM the choice regarding the number of harmonics used in an analysis is based on a trade off between the solution accuracy desired and the computer resources required.e. As noted in Section 12. and the solution accuracy would increase with the number of nodes used. These nodal values are expressed as a Fourier series which has the same number of coefficients as there are nodes. For the special case of a linear elastic continuous FSAFEM. where: a = 2TT//7V. The continuous FSAFEM uses a 2D axi-symmetric mesh and represents the variation of nodal displacements and loads in the circumferential direction as a continuous Fourier series. in general these displacements would not represent the correct displacement field exactly.4.3 for the continuous FSAFEM. The discrete method uses a discrete Fourier series to represent nodal displacements and is able to exactly represent any variation of nodal displacements in the circumferential direction. / is an integer and 0 < / < ns-1. It is assumed that the stresses at the sampling points are correct. i. Only the variation of radial stress around the pile (0 direction) is studied and the variation in the radial or vertical direction is not considered. ns is equal to the number of sampling points. To assess the accuracy of the two FSAFEM. their ability to represent the distribution of radial stress around a pile is considered. The stress distribution around a laterally loaded pile was presented by Williams and Parry (1984). the number of harmonics required in an analysis is equal to the number of harmonics required to represent the boundary conditions. the number of sampling points in the 6 direction is equal to the number of harmonic coefficients used in the FSAFEM analysis. The accuracy of this representation depends on the number of harmonic coefficients used in an analysis. However. Increasing the number of coefficients in a nonlinear discrete FSAFEM analysis increases the number of nodes in the 6 direction and increases the solution accuracy. In a discrete FSAFEM the distribution of nodal variables in the 6 direction is represented using the nodal values and shape functions of the appropriate annular ring of 3D brick elements. thus the problem is one dimensional. In particular. The sampling points are equispaced in the 6 direction and are located at 9 coordinates ia. using measurements from a model pile test. they are equal to the measured stresses shown in Figure 12.

Continuous In a discrete FSAFEM analysis with constant strain elements. It is assumed that these stresses represent the stress state at the centre of the element. the elements have a constant width in the 6 direction. Similarly. This error. since these elements have mid-side nodes. It is assumed that the stresses at the end nodes are correct and there is a linear stress distribution between them. The value of Err associated with each type of analysis is presented in Table 12. Due to the stress distribution.15: Postulated and measured coordinates ia. but could be in error 20-node discrete between sampling points. The stresses at the mid-side node depend only on the stresses at the end nodes. the number of nodes is also equal to n and they are located at 6 Figure 12.Fourier series aided finite element method / 393 70 representation for radial stress Measured stress is correct at the sampling 8-node discrete points. F(6) = the postulated stress distribution. there would now be In nodes. However. The difference between the desired stress and the stress from these analyses is a measure of the error associated with these analyses. If the number of 135 90 Azimuth (deg) coefficients is equal to n. . Due to the quadratic nature of the displacements. the number of elements is also equal to n.90) where: f{6) = the measured stress distribution. The stress distribution from each of the three types of analyses are presented in Figure 12. Err.15. for n equal to ten. the stresses and strains within an element are constant. For any value of n the continuous method and linear strain discrete method have similar accuracies and the constant strain discrete method is substantially less accurate. the stresses and strains within an element are linear. This is similar to the assumption made regarding the sampling point stresses for a continuous FSAFEM analysis. However. the elements still have a constant width and the nodes are still equispaced in the 6 direction. there are n unknown harmonic coefficients in the continuous method and In unknown nodal values in the linear strain discrete method. In a discrete FSAFEM analysis with linear strain elements. if the number of coefficients in a discrete FSAFEM analysis is equal to n. Hence. n=1O linear nature of the displacements.2. is quantified using the equation: I yj(F(O)-f(0))2 dO Err = • •100% (12.

7 % 19.8 % Some authors (e.5 % Discrete Fourier series using 20 node elements 48. the magnitude of the maximum -1 0 error. -1 0 However.16 for a square wave.8 % 7. This is illustrated in Figure 12.g.8 % Continuous Fourier series 47.5 % 21.2: Summary of errors in the representation of radial stress around pile Number of harmonic terms 2 5 8 10 Discrete Fourier series using 8 node elements 100. i.16: Fourier series error in the vicinity of the discontinuity. Table 12. the Gibb's phenomenon'.5 % 17.g.394 / Finite element analysis in geotechnical engineering: Theory the stiffness matrix associated with the linear strain discrete method is larger than for the corresponding continuous method. This would result in increased computer resources required for an analysis. while increasing the n is the number in the Fourier series number of Fourier terms does not reduce the magnitude of Err.1 0 1 x equal to 0. The error Err is approximately . The Gibb's phenomenon states that if a Fourier series is used to represent a function which has a discontinuity. statements like this are ambiguous since the Gibb's phenomenon does not suggest such a problem.e. Err. However. The Gibb's phenomenon would only be relevant in a finite element analysis if a discontinuity in displacements existed in the circumferential direction. Thus it is concluded that the continuous FSAFEM is probably the most efficient method. Lai and Booker (1991)) have stated that the continuous Err\ FSAFEM could suffer from 'difficulties associated with summing a large number of Fourier terms.9% 10. of the Fourier series representation in the vicinity of the discontinuity is largely Err\ 77 = 3 2 independent of the number of Fourier terms used. and this is not feasible in .0% 30. e.4 % 17.09 times the magnitude of the of harmonic terms used discontinuity.9 % 14. approximation of a square the area between the Fourier series and the wave square wave is reduced.2% 6. it does reduce the total Figure 12.

Each set of analyses consisted of a number of separate analyses performed with a different number of Fourier harmonics.17.g. However.2.2.• Smooth . The displacements in the 6 direction for a continuous FSAFEM analysis are described by a continuous Fourier series. G. It would only be possible if non axisymmetric interface elements were present.4. Thus at the boundary between the pseudo 2D continuous Fourier elements and the 3D elements. against number of harmonics used in the analysis. if isoparametric elements are used. Two sets of analyses were performed. This criticism of the continuous FSAFEM is therefor not valid.7. butmodelled the soil as a Tresca material. the displacements are equal at the nodal points of the 3D mesh.Fourier series aided finite element method / 395 a conventional finite element analysis.• .CFSAFEM Rough . In this case the displacements between nodes at the boundary. Pu I StlD (where Pu is the ultimate load and D is the diameter of the disk).. of 30 kPa. are identical. Su. Lai and Booker (1991) have also stated that 'problems concerning conformity of elements is overcome' by using the discrete method rather than the continuous method. number of conditions as described in harmonics for a pile section Section 12.DFSAFEM 2 0 " 0 2 1 4 1 6 8 1 0 1 2 u 1 6 Number of hannonic coefficients 1 1 1 1 1 1 1 finite element code (ICFEP). can be made using results for the horizontal disk problem described in Section 12. e. of 3000 kPa. and of the influence of the number of Fourier coefficients employed in each method. the other a rough. //.17: Limit load vs. which is between the pseudo 2D discrete Fourier elements and the 3D elements. His results are also shown in Figure 12.4. but do not match at intermediate values of 6. and it is unclear how either the discrete FSAFEM or continuous FSAFEM could incorporate such behaviour. The . Ganendra (1993) performed continuous FSAFEM analyses of this problem using the Authors' 12 0 © Analytical rough D 9? n > / _ ^ # H • C 8 Analytical smooth 6 i\ >v / M -e- Smooth . one assuming a smooth. Lai (1989) also analysed this same problem. He used the same boundary Figure 12.CFSAFEM Rough . with a shear modulus. and thus they match. a Poisson's ratio. but used the discrete FSAFEM. This statement is not clear: the continuous method should have no element conformity problems since the displacements at a boundary. The results are presented in Figure 12. both methods can be adapted to undertake an analysis consisting of both full 3D elements and pseudo 2D FSAFEM elements. are described by both types of elements using the finite element approximation. a core consisting of full 3D elements surrounded by layers of pseudo 2D elements.DFSAFEM . A further comparison between the continuous and discrete FSAFEM.17 in the form of normalised ultimate load.7.49 and an undrained shear strength. equal 0. interface between rigid disk and soil. calculated from either element sharing that same boundary.

see a) Horizontal load on a circular footing Figure 12. the CFSAFEM analysis used a 2D finite Figure 12.g. Inspection of Figure 12.19a and b) Inclined load on a circular footing 12.17 indicates that the continuous FSAFEM is not strongly influenced by the number of harmonics used.396 / Finite element analysis in geotechnical engineering: Theory analytical solutions for the limit loads for the rough and smooth interface cases are 11.HSJD respectively (Randolph and Houlsby (1984)).18. The soil is assumed to P behave according to a Tresca yield criterion. To be consistent. The discrete method is very sensitive to the number of harmonics employed.7 Comparison of CFSAFEM and full 3D analysis To investigate the benefits of the CFSAFEM over full 3D analysis.945SUD and 9. using both approaches. The rough solution is exact. a rough rigid circular footing under horizontal and inclined loading has been analysed.18: Loading schemes for circular footing element mesh which was the same as the vertical section of the 3D mesh shown in Figure 12. and if less than ten harmonics are used significant errors are obtained. the CFSAFEM was run using the parallel symmetry option. 12. and reasonable accuracy is obtained if five or more harmonic terms are used. e.19a: Horizontal cross-section of a 3D mesh for a circular footing . ju=0A5 and $=100 kPa. with E= 10000 kPa. Due to the symmetry mentioned above. only one half of the problem requires analysis.19b. the error is greater than 65% if 6 harmonics are used. while the smooth solution is a lower bound.19b. 100 m if & Figure 12. As there is a vertical plane of symmetry. These results are indicative of the number of harmonics required to analyse a laterally loaded pile accurately and suggest that (i) the continuous method is more economical in terms of number of harmonics required and (ii) large errors can be obtained using the discrete method if an insufficient number of harmonics are used. The mesh for the 3D analysis is shown in Figures 12.

03 0. the CFSAFEM with five harmonics required approximately a tenth of the time of the full 3D analysis. expressed in terms of the horizontal force on the footing versus horizontal displacement. are given.19b: Vertical cross-section of a 3D mesh for a circular footing.05 Horizontal displacement (m) 0.Fourier series aided finite element method / 397 Applied load a> 50 m <B Figure 12.20. It can be seen Figure 12. However.02 0. are given in Figure 12.04 0. Also noted on the figure are the run times CFSAFEM ! 3000 - 3D 5 harmonics 10 harmonics 20 harmonics (41 hours) (4. also a mesh for CFSAFEM analysis The footing is displaced horizontally and the horizontal reaction on the foundation noted. For example. Results from a full 3D analysis and three CFSAFEM analyses.5 hours) (7 hours) (21 hour) o.20: Load-displacement curves that all analyses produce very for a horizontally loaded circular footing similar results. each with a different number of harmonics.06 0. As the displacement .oi 0.07 for each analysis. the CFSAFEM have much smaller run times. at an angle of 45° to the horizontal. A second series of analyses was then performed in which the footing was displaced downwards. with a maximum difference of only 1%. The results from the analyses.

01 0. as opposed to parallel or orthogonal symmetry analysis. As with for a circular footing under inclined the horizontal loading.398 / Finite element analysis in geotechnical engineering: Theory increased.21. in the form of horizontal 3D (55 hours) load plotted against horizontal displacement and vertical load plotted against vertical displacement. The continuous Fourier series aided finite element method (CFSAFEM) has been described in detail in this chapter. . 4. The definitions of.06 0. both the horizontal and vertical loads on the foundation were recorded. and vertical. 3. The results are shown in Figure CFSAFEM (18 hours) 12. In this chapter the method is applied to problems having an axi-symmetric geometry. u. Results are given 0.04 0.8 Summary 1.05 0. the bulk compressibility of the pore fluid and coupled consolidation has been presented. while displacements are assumed to vary according to a continuous Fourier series in the third dimension. are available within the computer code being used. symmetrical and harmonic coupling have been established and the associated computational savings identified.02 0. A new formulation for the CFSAFEM for nonlinear material behaviour has been described. In this approach a standard finite element discretisation is used in two dimensions. These two examples clearly show that the CFSAFEM is more economical than a full 3D analysis. 12. The definitions of parallel and orthogonal symmetry have been given and the requirements for such symmetries to be valid have been discussed. with displacements varying according to a Fourier series in the 6 direction. It is advised that facilities to undertake no symmetry. A 2D finite element discretisation is used in the r-z plane. butnon axi-symmetric distribution of soil properties and/or loading conditions. and requirements for. v.07 for a full 3D analysis and for a Horizontal. Extension of the CFSAFEM to include interface elements. The run times are Figure 12. This method provides a way of reducing the computer resources required to analyse certain 3D problems.21: Load-displacement curves also noted on the figure. 2. It is shown how subtle changes in the constitutive behaviour can lead to violation of these requirements and hence to erroneous results. displacement (m) CFSAFEM analysis with 10 harmonics. both loading analyses give similar predictions and again the CFSAFEM is much faster.03 0. A general approach has been adopted and no symmetry constraints have been imposed. One of the requirements concerns the nature of the constitutive model.

4) Carrying out the integral and using the orthogonal properties of the Fourier series gives: AW = 2nr Ad°AF° + nrt k=\ Adk AFk + Adk AFk . The discrete method requires a full three dimensional finite element mesh defined in cylindrical coordinates. For the examples considered. The distribution of nodal variables in the 6 direction is represented using the nodal values and shape functions of the appropriate annular ring of 3D brick elements. Ad. Substituting the Fourier series into the work equation gives: AW= ) \Ad° + j^ Adk cosk0 + Adk sin*0| AF° + £ AFk cosk0+AFk smk0\r d0 -n V k=\ J\ k=\ J (XII. The theory behind this approach and the difference between the two methods have been discussed. An alternative to the continuous FSAFEM is the discrete FSAFEM. Appendix XII-1: Harmonic coefficients of force from harmonic point loads The applied incremental load. 7. at any angle 6 is written in the form of a Fourier series: AFk cosk0+AFk sinkO k=l (XII. AF (component of force per unit circumference). be described using a Fourier series: Ad = Ad° + S Adk coskO+ Adk s'mkO Thus the work done is: AW= ] AdAF r&O -71 (XII. each method has its advantages. These nodal values are expressed as a discrete Fourier series which has the same number of coefficients as there are nodes.Fourier series aided finite element method / 399 5. Comparison of the continuous and discrete FSAFEM indicates that the former method is likely to be more economical. Comparison of CFSAFEM and full 3D analysis shows that the former requires considerably less computer resources to obtain answers of the same accuracy. However.3) where r is the radius of the circle described by the location of the node. 6.2) (XII. savings in computer time are up to an order of magnitude. Let the associated component of incremental nodal displacement.

z and 6 directions to obtain ar. can be written as: Ad=t NAAd? + £ Adk cosk0+Adk sinko] (XII.2: Obtaining the harmonics of force from harmonic boundary stresses Boundary stresses are prescribed over an axi-symmetric surface defined by a section of the finite element mesh boundary. used in the right hand side of the FSAFEM equations. The work done can then be written as: AW =\ Ad ACT darea = j J Ad ACT rdOdl . with a cubic spline function. These stresses can be expressed as a Fourier series. rrz. Adf.6) Appendix XII. (XII. are: AT?0 = 2nr AF° ARk = nr AFk ARk =nrAFk for the 0th harmonic. These stresses can be resolved into global r. Thus for any component of incremental stress.400 / Finite element analysis in geotechnical engineering: Theory Using the principle of virtual work. a Fourier series can be written such that: ACT = ACT0 + £ Acrk coshO+ Acrk smkO (XII.9) where d/ is the incremental length in the r-z plane along the boundary. a0. Adk and Adk are the 0th. az. for the Ith node. for example. at any point in the r-z plane. and the variation of each harmonic coefficient along the boundary is described.8) where: Nj is the shape function of the 7th node defined in the mesh.7) The displacement over the axi-symmetric surface is described using the shape functions of the appropriate finite elements and Fourier series of the nodal displacement of the element. AR. Thus for Aa the corresponding component of incremental displacement. the kth cosine and the kth sine coefficients of incremental displacement respectively. rr0 and rz0 respectively. for the kth sine harmonic.7 t (XII. the contributions to the harmonic coefficients of incremental nodal force. Substituting the Fourier series representation in this equation gives: AW=\ J ]T NA A^° + ]T Adk coskO+Adk k sink0\\ ACT0 + £ ACT7 cosl0+Aal A /=1 sin/0 r d0d/ J (XII. n is the number of nodes in the element. 10) . Ad. ACT. for the kth cosine harmonic.

the Fourier series representation can be divided using these two subvectors. 16) .z}T and {Aa2} = {AarB Aa20}T Accordingly. i. {A of}. the kth cosine and the kth sine coefficients of stress respectively.3: Obtaining the harmonics of force from element stresses The stress point algorithm evaluates the incremental stresses at integration points within the problem domain. 11) J The d/ integral is carried out numerically. usually using a Gaussian integration rule. AR? = nr\ Ni Acrk d/ ARf = nr\ Nf A<j d/ k for the kth cosine harmonic. i.e: {Ao) = {Acrr Aa= Aa0 Aarz\Aar0 where: {Aoi} = {Ao-.e: {ACT} = {Aor0} + £ {A^} cos£<9+ {Ao*} smkO (XII. These stresses are expressed as a Fourier series.e: Aa=0}T = {{Acrl} {Acr2}}T (XII. ARf. 13) where: {A<r} is the vector of incremental stress components. are: AR° = 2nr\ Ni ACT0 d/ for the 0th harmonic. th (XII. {Acl} and {Ao"2}. Aaz Aae Acr. i. Vector {Aa} can be split into two sub-vectors. [ACT-} and {Aaf} are the vectors of the 0th. 12) Appendix XII. Thus using the principle of virtual work the contributions to the harmonic coefficients of incremental nodal force at the fh node. 14) [{Ao-2°}J *=' [{Ao-2'}cos^J [{Acr2*} sinifc^J The incremental internal work done by these stresses is: j {A^}T{Acr} dVol=\) {Ad}T[Bf {ACT} r&O&area (XII. for the k sine harmonic.Fourier series aided finite element method / 401 Carrying out the d6 integral gives: Ad* Ao* + Arf* ACT* I d/ (XII. used in the right hand side of FSAFEM equations.

20) l-[«']J [{A. the 8 integral is carried out: AW = %± -2n{Ad°**}T[B2°]T{Acr20} [[B2']\ [{A<r2'}J darea (XII. Equation (XII. [ ' J N ^ A ^_^ I r darea for the /th orthogonal harmonic. 16) can be expressed in a Fourier series form as: [Bl1] cos/0 /=o I -| T r ^ T [Bll] sin/0 ~[B2l] cos/0 i^ [B2l] sin/0 1 {A<x2 } 0 k L [{Acrr*}cosA. [-[S2°]J [{A<x2°}J {AR1*} =' 7t JI {A/?'"} = TCH {[ } 1 U {A Z=} 1 r darea for the /* parallel harmonic. From here the contributions to the harmonic coefficients of incremental nodal load. 18) The work done by the coefficients of incremental nodal force consistent with this stress state is: AW= {Ad°*}{AR°*}+t {Adk*}{ARk*} + {Adk**}{ARk**} (XII. (XII.402 / Finite element analysis in geotechnical engineering: Theory where: {Ad} is the vector of incremental displacements.2'}j . 19) Thus using the principle of virtual work. are: {AR°} = 2TIJ \ f [*l o ]] T f{A<xl o }l U \r darea for the 0th harmonic.0] -' { r dO darea (XII. the work done by both equations must be the same and in addition the terms which are pre-multiplied by the same coefficients of displacement can be equated. [B] is the strain matrix. 17) Using the orthogonal properties of the Fourier series. used in the right hand side of the FSAFEM equations. {AR}.

AT. ATr^Vzn. ATr°~°. This can be expressed as a Fourier series: Afr = AFr° + £ AFrk cosk0+ AFrk smk9 k=\ (XII. is: AT?*** = ) Aforcos0 d0 -71 = J f AF0° + t A F / cosk0 + AFk sinA:6>jcos6> r d6> -71 V (XII..25) n=l / = nr AFQ = ARxe Thus the total incremental horizontal load in the 0=0° direction is: ATre=0 = ARi-A^ and the total incremental horizontal load in the 6=l/2K direction is: °^ ~ (XII. ATrfh0.23) AF ° + t AFk cosk0+AFk sinA^lsin^ r d0 .Fourier series aided finite element method / 403 Appendix XII. the resultant incremental force in the O^Vin direction.26) . the resultant incremental force in the #=0° direction.22) = ) (AFr° + X AFk cosk0+AFrk sinik^lcos^ r d0 = nr AFrl = ARlr Similarly. _ r k± j = nr AFl = < fff represents the incremental circumferential force per unit circumference on a node. is given by equation: ATr0=0 = ] -n Afrrcos0d0 (XII. is: ATer=*n = ) Afrrsin0 d0 (XII.4: Horizontal loads Resolving harmonic coefficients of nodal force Let Afr represent the incremental radial force per unit circumference on a node. is given by: = ] -(AF0° + £ AFk cosk0+ AFg sinA:<9]sin(9 r d0 -n V k=l ) = .0=V2n.27) (XII. the resultant incremental force in the 0=l/2n direction.21) The resultant incremental force in the #=0° direction.n r AF e = l -ARXO Similarly.

AM/2%. This can be expressed as a Fourier series: Afz = AFZ° + t AFzk cosk0+ AFzk sinkO (XII. is equal to: AM*" = ] Af2rcos0 rd0 ^XH 30) = J \AFZ° + £ AFzk co$k6+AF2k smkO cos<9 r2 d6 = nr2AF2l =rARl2 Similarly. for incremental turning moments about the #=0° axis. -71 =0ifk = l=0 (XII. is given by the equation: AT2 = ] A = ) (AF2° + t Aff cosk0+ AF? sinker d9 = 2nrAF? = ( x n 29 ) Turning moments As in the previous case for axial load.404 / Finite element analysis in geotechnical engineering: Theory Axial loads Let Afz represent the incremental vertical force per unit circumference on a node. ATZ. let Af2 represent the incremental vertical force per unit circumference on a node. AAf is equal to: AM°= J Af2rsm6 rd6 -n = 1 ( AFZ° + t AFk cosk9+ AFk smko\sinO r2 dO = nr2AF2l=rARl2 (Xn 3 ^ Appendix XII.32) J coskOcosWdO =0if .28) The total incremental force in the vertical direction. which can be expressed as a Fourier series.5: Fourier series solutions for integrating the product of three Fourier series Three standard integrals are used for the solution of the integral of the product of two Fourier series: J smkOcoslOdO = 0 for all k and / -71 J sin£<9sin/<9d<9 = 0 i f * * / . The incremental turning moment about the 0=l/2K axis. = n i f k = l*0.

= 0 otherwise.. ±is + i f £ ..33) Appendix XII. using n discrete values of xf at locations 0f. |x| denotes absolute value of x... xn at 5W. at #.1)0+ cos(& +1)0) sinmO dO =0 J sin£<9cos/<9sinw<9 d0= ) ~(sin(£ .6: Obtaining coefficients for a stepwise linear distribution The variable x is described by a stepwise linear distribution. = V2 if | i +/| = m * 0 ...g.34) 27ttl i . .. is . = lA if | k -l\ = m * 0 . (XII.1)0+ cos(£ +1)0) cosmO dO j sinkOcoslOcosmO dO= J —(s'm(k -1)0+ sin(£ +1)0) cosmO dO -n -n ^ =0 where: a = 1 if |* -l\ = m = 0 . = 0 otherwise.1)0+ sin(k +1)0) sinmO dO sinkOs'mlOcosmO dO= ] ~-( cos (^ ~ l)e~ cos k ( +1)°) ^osmO dO ] coskOcoslOcosmO d0=) —(cos(A../ > 0 . such that at any location 0 the value of x is: /) The 0th harmonic is evaluated from: when gj^ < g < ^ +i (XII./ < l .4. see Figure 12. xx at 0x. e.Fourier series aided finite element method / 405 Using the above equations six standard integrals can be obtained for the solution of the integral of the product of three Fourier series: J sinkOsinlOsinmO dO = ] ~(cos(k-l)0-cos(k -n -n *• + l)0) sinmO dO =0 ] coskOcoslOsinmO dO = f ^(cos(£ . x. P = 1 if \k +/| = jw = 0 . x2 at 62.i f * ..

.6!) .^)g+^/^. I .37) The kth sine harmonic is evaluated from: Xk =— Y f(^.406 / Finite element analysis in geotechnical engineering: Theory 2 1 A */+i+*.-cosA.„ . the summation of the last two terms in the expression is zero. . -cosM # . giving: (XII.7+1 = ^ j .35) The kth cosine harmonic is evaluated from: Xk =— J xcoskO dO /=i v ^.-ffl>i .)| sin ^ /+1 Jsinkd 0cosk0) ^ \ ~ F " ~ST"J . f/1 1 ^ ^-^ r /+1 sin^ +1 -x7sin^6!j ( X IL36) Since xn+l = xl and #. Of 0 0 = THh iAi+l ~ i} (XII.

'0 Figure XII. For each value of x specified.2.Fourier series aided finite element method / 407 (xn38) The summation of the last two terms in the expression is zero. for each x. a symmetrical value is assumed in the half space -n < 6 < 0. to x. at -#. Since the assumed distribution is asymmetric. n is . For each value of x specified. an asymmetrical value is assumed in the half space -n<0 <0.e. Specified X Assumed *«-• M N x2 x Specified xx \ \ . the sine coefficients are zero. In order to have the same number of sine and cosine harmonics.39) Simplifications can be made if x varies in either a symmetric or non-symmetric manner. see Figure XII. the 0th and cosine coefficients are zero. see Figure XII. is assumed. for each x. 1. the specified values of x are constrained to be located in half space 0 < 6 < n. at #.). 2: Asymmetrical stepwise linear function Similarly.e.. i. Appendix XII. The above equations can then be used to obtain 0th and cosine harmonic coefficients. a value of -x.7: Obtaining harmonic coefficients using the fitted method A Fourier series has to be found which passes through n discrete values of x (x. giving: k = (XII. Since the assumed distribution is symmetric. The above equations can then be used to obtain the sine harmonic coefficients. at 0f a value of X. the specified values of x are constrained to be located in the half space 0 < 6 < it. if an asymmetric function is stipulated. If a symmetric function is stipulated. i. at -6} is assumed. 1: Symmetrical step wise linear function Figure XII.

44) Equation (XII.1) — f. which gives: .6. JC.1) coska(j .41) The harmonic coefficients for this Fourier series can then be evaluated using the equations: n k /=i X =.43) can be simplified: JC. i. = X° + £ Xk cosk{j-\)a+Xk smk(j-\)a k=] (XII. + . L.=-± xl\ + t coska(i .1) sinka(j-l) n JI] n Noting that cosA cos^ + sin^ sinB = cos(A . sin*ar(/. . coska(i . Thus the order of the Fourier series required. iffy where: ¥ij (XII.e.42) into Equation (XII.t xt sink(i .408 / Finite element analysis in geotechnical engineering: Theory constrained to be an odd number. From here Equation (XII.42) This can be proved by substituting Equations (XII.S x.=i k (XII.t x.l)a.41).cosk(i-l)a n i=\ X = .45) A parameter y/^ can now be defined such that Equation (XII.B) Equation (XII. x}is located at 6 = (j.j)) (XII.45) becomes: Xj= .40) The values of x are equispaced in the region -n < 6 < n.44) can be rewritten so that the summation of A : is carried out first: x.\)a n .i t xi coska(i .j) (XII.46) 2 k={ cos a l J . is equal to !/£(«-1): x = X° + £ Xk cos£6>+ Xk sin£<9 k=\ (XII.40) can be written such that for any integer value j : x.t x.. see Figure 12. = — i x .

j = (L+V£) since the cosine of zero is 1.. Equation (XII. 3: Symmetrical function Figure XII.46) can be written as: (XII.. = x.Fourier series aided finite element method / 409 Using the standard solution: 1 and if/ * j then: L — +Ya COSAX = 2 k=\ 2sin- T (XII. i. This is illustrated in Figure XII.3.42) yields harmonic coefficients for a Fourier series which satisfies Equation (XII. \fi=j.47) " 2sin^il Substituting L into the equation for a gives: a _ 2TT = 2TI n 2L+1 Substituting this into Equation (XII. Simplifications can again be made if x varies in either a symmetric or nonsymmetric manner. 4: Asymmetrical function .48) Equation (XII. the sine coefficients are zero. then values of x are only specified in the half space -7t<#<0. Since the assumed distribution is symmetric. and the sine of any multiple of n is zero. Equation (XII. If a symmetric function is stipulated.e.48) equals zero for / *j because (j~i) is an integer._. Assumed Assumed Specified Specified Figure XII.49) Thus Equation (XII.47) for ^ gives: w _ smn(j-0 _n (XII. then y/.41) for all values ofy. only x0 to xL is specified and xIM to xn is obtained using the equation x.42) can now be used to obtain the zeroth and cosine harmonic coefficients.

.42) can now be used to obtain the sine harmonic coefficients. if an asymmetric function is stipulated. then values of x are only specified in the half space 0<0<n and an asymmetrical value is assumed in the half space -7r<#<0.410 / Finite element analysis in geotechnical engineering: Theory Similarly.e. i. the zeroth and cosine coefficients are zero. = -xn.4. This is illustrated in Figure XII. Equation (XII. only xx to xL is specified and xL+l to xn is obtained using the equation x.f. Since the assumed distribution is asymmetric. x0 is constrained to be zero for an asymmetric function.

Sign up to vote on this title
UsefulNot useful