Você está na página 1de 25

# 12

A Complete
Axisymmetric
FEM Program

12–1
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–2

Page
§12.1. ANALYSIS STAGES 12–3
§12.2. PROBLEM DEFINITION 12–3
§12.2.1. Illustrative Problem . . . . . . . . . . . . . . . 12–4
§12.2.2. Node Coordinates . . . . . . . . . . . . . . . 12–4
§12.2.3. Element Type . . . . . . . . . . . . . . . . . 12–5
§12.2.4. Element Connectivity . . . . . . . . . . . . . . 12–6
§12.2.5. Material Properties . . . . . . . . . . . . . . . 12–6
§12.2.6. Fabrication Properties . . . . . . . . . . . . . . 12–6
§12.2.7. Freedom Tags . . . . . . . . . . . . . . . . . 12–7
§12.2.8. Freedom Values . . . . . . . . . . . . . . . . 12–7
§12.2.9. Element Processing Options . . . . . . . . . . . . 12–7
§12.2.10.Element Body Forces . . . . . . . . . . . . . . 12–8
§12.2.11.Mesh Display . . . . . . . . . . . . . . . . . 12–8
§12.3. PROCESSING 12–9
§12.4. POSTPROCESSING 12–10
§12.4.1. Printing . . . . . . . . . . . . . . . . . . . 12–10
§12.4.2. Displacement Field Contour Plots . . . . . . . . . . 12–10
§12.4.3. Stress Field Contour Plots . . . . . . . . . . . . . 12–11
§12.5. EXAMPLE 1: INTERNALLY PRESSURIZE THICK TUBE 12–14
§12.6. EXAMPLE 2: ROTATING THIN DISK 12–16
§12.7. EXAMPLE 3: CENTRALLY LOADED SS CIRCULAR PLATE 12–20
EXERCISES . . . . . . . . . . . . . . . . . . . . . . 12–24

12–2
12–3 §12.2 PROBLEM DEFINITION

## This chapter describes a complete axisymmetric FEM program, implemented in Mathematica

Notebook Quad4.SOR.nb. Unlike the previous chapter, the description is top down, i.e. starts
from the main driver programs. Three benchmark examples are included to illustrate the operation
of the code and the level of accuracy attainable with 4-node isoparametric ring elements.

## §12.1. ANALYSIS STAGES

As in all FEM programs, the analysis of a structure by the Direct Stiffness Method involves three
major stages: (I) preprocessing or model deﬁnition, (II) processing, and (III) postprocessing.
The preprocessing portion of the axisymmetric analysis is done by the driver program, which carries
out two activities:
I.1 Model deﬁnition by direct setting of the data structures.
I.2 Plot of the FEM mesh, including nodes and element labels.
The processing stage involves three steps:
II.1 Assembly of the master stiffness matrix and force vector, using the element library.
II.2 Application of displacement BC. This is done by the equation modiﬁcation method.
II.3 Solution of the modiﬁed equations for displacements. The built in Mathematica function
LinearSolve is used.
Upon executing these three processing steps, the displacements are available. Three postprocessing
steps follow:
III.1 Recovery of forces including reactions, carried out element by element.
III.2 Computation of element stresses and interelement averaging to get nodal stresses.
III.3 Printing results and selective plotting of displacement and stress ﬁelds.
The input data to a ﬁnite element program consists generally of three sets, which are associated
with three phases of the program execution:

##  Preprocessing phase: problem deﬁnition data

Input data Processing phase: problem solving data
Postprocessing phase: result display data

## §12.2. PROBLEM DEFINITION

The problem-deﬁnition data may be broken down into three sets, which are listed below by order
of appearance:

##  Geometry data: node coordinates

Model deﬁnition Element data: type, connectivity, material and fabrication
Degree of freedom data: forces and support BCs

Note that the element data is broken down into four subsets: type, connnectivity, material and
fabrication, each of which has its own data structure. (For axisymmetric solid models, the fabrication

12–3
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–4

z
2b
2a

TUBE CROSS r
SECTION p

internal pressure p
tube extends indefinitely
along the z axis

## Figure 12.1. A thick circular tube under internal pressure.

data is empty.) The degree of freedom data is broken into two subsets: tag and value. In addition
there are miscellaneous element process options, which are conveniently collected in a separate
data set.
Accordingly, the model-deﬁnition input to the axisymmetric stress FEM program consists
of eight data structures, which are called NodeCoordinates, ElemTypes, ElemNodeLists,
ElemMaterial, ElemFabrication, FreedomTags, FreedomValues and ElemOptions
These data sets are described in the following subsections with reference to the problems and
discretization of Figure 12.1–2.

## §12.2.1. Illustrative Problem

The problem illustrated in Figure 12.1 is a standard benchmark for axisymmetric ﬁnite elements.
A thick circular tube of internal radius a and external radius b is subject to internal pressure p. The
z axis is placed as shown. The tube extendas indeﬁnitely along the z axis. The material is isotropic
with elastic modulus E and Poisson’s ratio ν.
The tube is cut with a normal-to-z slice of width h as depicted in Figure 12.2(a). A two element
discretization using two 4-node quadrilateral ring elements is constructed as shown in Figure 12.2(b).

## §12.2.2. Node Coordinates

The geometry data is speciﬁed through NodeCoordinates. This is a list of node coordinates
conﬁgured as
NodeCoordinates = { { r1 , y1 },{ r2 , y2 }, . . . { r N , y N } } (12.1)
where N is the number of nodes. Coordinate values should be ﬂoating point numbers; use the N
function to insure that property if necessary. Nodes must be numbered consecutively starting from
1. No gaps are permitted.
The the model of Figure 12.2, with a = 4, b = 10 and h = 2:

12–4
12–5 §12.2 PROBLEM DEFINITION

## z (a) THICK TUBE UNDER

INTERNAL PRESSURE

z
(b) FEM DISCRETIZATION

lice"
h mi S

;;
;;;;
;;;;
;;
;;;;
"Sala
pr2
2 4
h 2 (1) 4 (2) 6 6
r

;;

;;
1 3 5 5
pr1 1 3
r=a
r = (a+b)/2
r=b

Figure 12.2. A slice of width h is cut from the thick tube of Figure 12.1 and used to
build a 2-element discretization with 4-node quadrilateral ring elements.

## a=4; b=10; h=2;

NodeCoordinates=N[{{a,0},{a,h},{(a+b)/2,0},{(a+b)/2,h},{b,0},{b,h}}];
The advantage of using this parametric form is that other values for the dimensions can be rapidly
tested.
The coordinates can be also constructed through mesh generation modules. One module provided
in the program accompanying this Chapter is GenerateRingNodeCoordinates, which is invoked
as follows:
a=4; b=10; h=2;
MeshCorners=N[{{a,0},{a,h},{b,0},{b,h}}];
NodeCoordinates=GenerateRingNodeCoordinates[MeshCorners,2,1];
The ﬁrst argument are the corner coordinates of a 4-sided region over which the mesh will be
generated, using a regular isoparametric mapping method. The second and third arguments are the
number of elements in the r and z directions, respectively. Generation is advantageous for exploring
mesh reﬁnement.

## §12.2.3. Element Type

Element type is a label that speciﬁes the type of element to be used.

## ElemTypes = { { etyp(1) }, { etyp(2) }, . . . { etyp(Ne ) } } (12.2)

Here etyp(e) is the type descriptor of the e-th element, speciﬁed as a character string. The
only element available in the axisymmetric program is the 4-node isoparametric quadrilateral ring
element identiﬁed as "Quad4". Thus for the example of Figure 12.2:
In practice a slightly different form is recommended:

12–5
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–6

Here numele is the number of elements. This can be extracted as numele=Length[ElemNodeLists],
where ElemNodeLists is deﬁned below.

## §12.2.4. Element Connectivity

Element connectivity information speciﬁes how the elements are connected.1 This information is
stored in ElemNodeLists, which is a list of element nodelists:

## ElemNodeLists = { { eNL(1) }, { eNL(2) }, . . . { enL(Ne ) } } (12.3)

where eNL(e) denotes the lists of nodes of the element e (given by global node numbers) and Ne is
the total number of elements.
Element boundaries must be traversed counterclockwise but you can start at any corner. Number-
ing elements with midnodes requires more care: ﬁrst list corners counterclockwise, followed by
midpoints (ﬁrst midpoint is the one that follows ﬁrst corner when going counterclockwise). When
elements have an interior node, as in the 9-node biquadratic quadrilateral, that node goes last.
For the mesh of Figure 12.2:

ElemNodeLists= {{1,3,4,2},{3,5,6,4}};
This data can also be constructed by generation modules if the mesh is sufﬁciently regu-
lar. A companion of GenerateRingNodeCoordinates that generates 4-node quadrilaterals is
GenerateQuad4NodeNumbers, which is invoked as follows:

The ﬁrst and second argument are the number of elements along r and z, respectively. These must be
exactly the same as the second and third arguments used with GenerateRingNodeCoordinates.

## §12.2.5. Material Properties

Data structure ElemMaterial lists the elastic modulus E and the Poisson’s ratio ν for each element:

## ElemMaterial = { { E (1) , ν (1) }, { E (2) , ν (2) }, . . . { E (Ne ) , ν (Ne ) } } (12.4)

In the frequent case in which all elements have the same E and ν, this list can be easily generated
by a Table instruction. For the mesh of Figure 12.2, assuming E = 1000 and ν = 0:
Em=1000; nu=0;
ElemMaterial=[{Em,nu},{numele}]
in which numele=Length[ElemeNodeLists] is the number of elements (2 there). Again it is
advantageoous to declare the numerical values separately for ease of modiﬁcation.

1 Some FEM programs call this the “topology data,” which sounds impressive.

12–6
12–7 §12.2 PROBLEM DEFINITION

## §12.2.6. Fabrication Properties

For axisymmetric solid models, this set is empty and should be set to the empty list:

ElemFabrication = { }; (12.5)

## §12.2.7. Freedom Tags

FreedomTags labels each node degree of freedom as to whether the load or the displacement is
speciﬁed. The conﬁguration of this list is similar to that of NodeCoordinates:

## FreedomTags={ { tagr 1 , tagz1 },{ tagr 2 , tagz2 }, . . . { tagr N , tagz N } } (12.6)

The tag value is 0 if the force is speciﬁed and 1 if the displacement is speciﬁed. When there are a
lot of nodes and comparatively few displacement BCs, the quickest way to build this list is to start
from all zeros, and then insert the boundary conditions appropriately.
For the mesh of Figure 12.2, all nodes happen to have the same condition: free to move along r ,
cannot move along z. One Table statement is sufﬁcient:
FreedomTags=Table[{0,1},{6}];
Rather than entering the number of nodes (6) directly, parametrization is recommended:
FreedomTags=Table[{0,1},{numnod}];
where numnod=Length[NodeCoordinates].

## FreedomValues={ { valr 1 , valz1 },{ valr 2 , valz2 }, . . . { valr N , valz N } } (12.7)

This array lists the speciﬁed values of the applied node force component if the corresponding tag is
zero, and of the prescribed displacement component if the tag is one. Often most of the list entries
are zero.
For the mesh of Figure 12.2, only two force values are nonzero: the r forces on nodes 1 and 2, dues
to the internal pressure. It is convenient to let the program calculate all values. For a = 4, p = 10
and h = 2:
a=4; p=10;
numnod=Length[NodeCoordinates];
FreedomValues=Table[{0,0},{numnod}];
FreedomValues[[1]]=FreedomValues[[2]]={Kfac*p*a*h/2,0};
Here Kfac is one of the element processing options.

12–7
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–8

RingSolution[nodcoor_,eletyp_,elenod_,elemat_,elefab_,
elebfor_,eleopt_,doftag_,dofval_]:= Module[{numnod,
u,dis,f,for,sig}, u=f=sig={}; numnod=Length[nodcoor];
u= RingAssemblyAndSolve[nodcoor,eletyp,elenod,elemat,elefab,
elebfor,eleopt,doftag,dofval]; u=Chop[u];
f= RingNodeForces[nodcoor,eletyp,elenod,elemat,elefab,
eleopt,u]; f=Chop[f];
sig= RingNodalStresses[nodcoor,eletyp,elenod,elemat,elefab,
eleopt,u]; sig=Chop[sig];
dis=Table[{u[[2*n-1]],u[[2*n]]},{n,1,numnod}];
for=Table[{f[[2*n-1]],f[[2*n]]},{n,1,numnod}];
Return[{dis,for,sig}];
];

Figure 12.3. Module to carry out the analysis of the axisymmetric problem.

## §12.2.9. Element Processing Options

Array ElemOptions should be set as follows:

ProcessingOptions={True,rule,Kfac};
The ﬁrst value requests ﬂoating point numerical computations for speed. The second value
is presently ignored. The third value is Kfac=1 if the ring element subtends one radian or
Kfac=N[2*Pi] if the ring element subtends a complete circle of 360◦ = 2π radians.

## §12.2.10. Element Body Forces

If body forces such as gravity or centrifugal forces are considered, an array ElemBodyForces must
be set up. For details see Example 2 in §12.4. If body forces are zero, set this array to the empty
list:
ElemBodyForces = { }; (12.8)

## §12.2.11. Mesh Display

Nodes and elements of may be plotted by the following statement:

Plot2DElementsAndNodes[NodeCoordinates,ElemNodeLists,aspect,
title,True,True];
Here aspect is the plot frame aspect ratio (y dimension over x dimension), title is a character
string specifying a plot title, and the last two True argument values specify node labels and element
labels, respectively. Such polots may be seen in the benchmark example sections.

12–8
12–9 §12.3 PROCESSING

RingAssemblyAndSolve[nodcoor_,eletyp_,elenod_,elemat_,elefab_,
elebfor_,eleopt_,doftag_,dofval_]:= Module[{i,j,ii,jj,n,
ne,ns,numer,rule,Kfac,zero=0,one=1,type,Emat,eNL,ncoor,
eftab,Ke, numnod=Length[nodcoor],numele=Length[elenod],
numdof,neldof,supdof,supval,dk,fixed,dfac=0,tag,f,fv,fb,
K,u,modname="RingAssemblyAndSolve: "}, numdof=2*numnod;
{numer,rule,Kfac}=eleopt; If [numer, zero=0.0; one=1.0];
K=Table[zero,{numdof},{numdof}]; u=fb=Table[zero,{numdof}];
For [ne=1,ne<=numele,ne++,
{type}=eletyp[[ne]];
(*If [type!="Trig3"&&type!="Trig6"&&type!="Trig10"&&
"Illegal element type,ne=",ne]; Return[Null]];
eNL=elenod[[ne]]; n=Length[eNL];
eftab=Flatten[Table[{2*eNL[[i]]-1,2*eNL[[i]]},{i,n}]];
ncoor=Table[nodcoor[[eNL[[i]]]],{i,n}];
{Em,ν}=elemat[[ne]]; Emat=Em/((1+ν)*(1-2*ν))*
{{1-ν,ν,ν,0},{ν,1-ν,ν,0},{ν,ν,1-ν,0},{0,0,0,1/2-ν}};
(* If [type=="Trig3", Ke=Trig3IsoPRingStiffness[ncoor,
Emat,Kfac,{numer}] ]; *)
Emat,Kfac,{numer,2}] ];
(* If [type=="Trig6", Ke=Trig6IsoPRingStiffness[ncoor,
Emat,Kfac,{numer,3}] ]; *)
Emat,Kfac,{numer,3}] ]; *)
(* If [type=="Trig10",Ke=Trig10IsoPRingStiffness[ncoor,
Emat,Kfac,{numer,3}] ]; *)
Ke=Chop[Ke]; neldof=Length[Ke];
For [i=1,i<=neldof,i++,ii=eftab[[i]];
For [j=i,j<=neldof,j++,jj=eftab[[j]];
K[[jj,ii]]=K[[ii,jj]]+=Ke[[i,j]]
];
];
];

## Figure 12.4. First part of the RingAssembleAndSolve module.

§12.3. PROCESSING

The static solution is carried out by calling module RingSolution shown in Figure 12.5. The
function call is
{NodeDisplacements,NodeForces,NodeStresses}=RingSolution[
NodeCoordinates,ElemTypes,ElemNodeLists,
ElemMaterial,ElemFabrication,ElemBodyForces,
ElemOptions,FreedomTags,FreedomValues];
Module RingSolution assembles the master stiffness matrix, constructs the right hand side,
applies the displacement boundary conditions, and solves for displacements by calling module
RingAssemblyAndSolve. This module is listed in Figures 12.4 and 12.5. As a study of its code
reveals, it can handle several element types but only "Quad4" is presently enabled. The modules
that compute the element stiffness matrices have been studied in previous Chapters.

12–9
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–10

(*Print["K=",K//MatrixForm]; *)
(*Print["eigs of K=",Chop[Eigenvalues[N[K]]]];*)
If [Length[elebfor]>0,
fb=RingBodyForces[nodcoor,eletyp,elenod,elebfor,eleopt]];
supdof=supval={}; fixed=Table[False,{numdof}];
For [n=1,n<=numnod,n++, For [j=1,j<=2,j++, i=2*(n-1)+j;
dfac=Max[dfac,K[[i,i]]]; tag=doftag[[n,j]];
If [tag>0, fixed[[i]]=True; AppendTo[supdof,i];
AppendTo[supval,dofval[[n,j]]] ]]];
(*Print["supval=",supval]; Print["supdof=",supdof];*)
If [dfac==0 || !numer, dfac=one];
ns=Length[supdof]; fv=Flatten[dofval]; f=fv+fb;
For [k=1, k<=ns, k++, i=supdof[[k]]; dk=supval[[k]];
If [dk==0, Continue[]];
For [j=1, j<=n, j++, If [fixed[[j]],Continue[]];
f[[j]]-=K[[i,j]]*dk;
];
];
For [k=1, k<=ns, k++, i=supdof[[k]]; f[[i]]=dfac*supval[[k]];
For [j=1, j<=numdof, j++, K[[i,j]]=K[[j,i]]=zero];
K[[i,i]]=dfac*one;
];
(*Print["modified K=",K//MatrixForm]; Print["f=",f];
Print["eigs of Kmod=",Eigenvalues[N[K]]];*)
u=LinearSolve[K,f]; u=Chop[u];
Return[u];
];

## Figure 12.5. Second part of the RingAssembleAndSolve module.

The computation of the right hand side merges the effect of directly speciﬁed node forces and of
body forces speciﬁed element by element. To do the latter, module RingBodyForces, which is
listed in Figure 12.6, is used.
Displacement boundary conditions are applied through appropriate modiﬁcation of the master
stiffness matrix and of the right hand side.
The unknown node displacements u are then obtained through the built in LinearSolve function,
as u=LinearSolve[K,f]. This solution strategy is of course restricted to very small systems, but
it has the advantages of simplicity.
Finally, array sig contains the nodal stresses σr z , σzz , sigmaθ θ σr z at each node, recovered from the
displacement solution. This computation is driven by module MembraneNodeStresses, which
is listed in Figure 12.8. This computation is actually part of the postprocessing stage. It is not
described here since stress recovery is treated in more detail in a subsequent Chapter.

§12.4. POSTPROCESSING
Postprocessing are activities undertaken on return from membraneSoution. They include optional
printout of node displacements, node forces and node stresses, plus contour plots.
§12.4.1. Printing
Figure 12.9 gives a simple script for printing the three arrays returned by RingSolution.

12–10
12–11 §12.4 POSTPROCESSING

RingBodyForces[nodcoor_,eletyp_,elenod_,elebfor_,eleopt_]:=
Module[{numele=Length[elenod],numnod=Length[nodcoor],
numer,rule,ne,eNL,eftab,neldof,i,n,Em,ν,Emat,Kfac,ncoor,
fe,f,modname="RingBodyForces: "},
f=Table[0,{2*numnod}]; fe=Table[0,{8}]; numer=eleopt[[1]];
{numer,rule,Kfac}=eleopt;
For [ne=1,ne<=numele,ne++,
{type}=eletyp[[ne]];
Print[modname,"Illegal element type,ne=",ne]; Return[Null]];
eNL=elenod[[ne]]; n=Length[eNL];
eftab=Flatten[Table[{2*eNL[[i]]-1,2*eNL[[i]]},{i,n}]];
ncoor=Table[nodcoor[[eNL[[i]]]],{i,n}];
bfor=elebfor[[ne]];
(* If [type=="Trig3", fe=Trig3IsoPRingBodyForces[ncoor,
Emat,Kfac,{numer},be] ]; *)
Emat,Kfac,{numer,2},bfor] ];
(* If [type=="Trig6", fe=Trig6IsoPRingBodyForces[ncoor,
Emat,Kfac,{numer,3},be] ];
Emat,Kfac,{numer,3},be] ];
If [type=="Trig10",fe=Trig10IsoPRingBodyForces[ncoor,
Emat,Kfac,{numer,3},be] ]; *)
neldof=Length[Ke];
For [i=1,i<=neldof,i++,ii=eftab[[i]];
f[[ii]]+=fe[[i]]
];
];
Return[f];
];

Figure 12.6. Module to process element body forces into node forces.

## §12.4.2. Displacement Field Contour Plots

Here is a generic script to do contour plots of displacement ﬁelds:

## ur=uz=Table[0,{numnod}]; Nsub=8; (* cut to Nsub=4 over finer mesh *)

Do[ur[[n]]=u[[2*n-1]]; uz[[n]]=u[[2*n]],{n,1,numnod}];
urmax=uzmax=0;
Do[urmax=Max[Abs[ur[[n]]],urmax]; uzmax=Max[Abs[uz[[n]]],uzmax],
{n,1,numnod}];
ContourPlotNodeFuncOver2DMesh[NodeCoordinates,ElemNodeLists,ur,
urmax,Nsub,aspect,"Displacement component ur"];
ContourPlotNodeFuncOver2DMesh[NodeCoordinates,ElemNodeLists,uz,
uzmax,Nsub,aspect,"Displacement component uz"];

## §12.4.3. Stress Field Contour Plots

Here is a generic script to do contour plots of stress ﬁelds:

12–11
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–12

RingNodeForces[nodcoor_,eletyp_,elenod_,elemat_,elefab_,
eleopt_,u_]:= Module[{i,j,ii,jj,n,
ne,numer,rule,Kfac,zero=0,one=1,type,Emat,eNL,ncoor,
eftab,Ke, numnod=Length[nodcoor],numele=Length[elenod],
numdof,neldof,f,modname="RingNodeForces: "}, numdof=2*numnod;
{numer,rule,Kfac}=eleopt; If [numer, zero=0.0; one=1.0];
f=Table[zero,{numdof}];
For [ne=1,ne<=numele,ne++,
{type}=eletyp[[ne]];
"Illegal element type,ne=",ne]; Return[Null]];
eNL=elenod[[ne]]; n=Length[eNL];
eftab=Flatten[Table[{2*eNL[[i]]-1,2*eNL[[i]]},{i,n}]];
ncoor=Table[nodcoor[[eNL[[i]]]],{i,n}];
{Em,ν}=elemat[[ne]]; Emat=Em/((1+ν)*(1-2*ν))*
{{1-ν,ν,ν,0},{ν,1-ν,ν,0},{ν,ν,1-ν,0},{0,0,0,1/2-ν}};
(* If [type=="Trig3", Ke=Trig3IsoPRingStiffness[ncoor,
Emat,Kfac,{numer}] ]; *)
Emat,Kfac,{numer,2}] ];
(* If [type=="Trig6", Ke=Trig6IsoPRingStiffness[ncoor,
Emat,Kfac,{numer,3}] ]; *)
Emat,Kfac,{numer,3}] ]; *)
(* If [type=="Trig10",Ke=Trig10IsoPRingStiffness[ncoor,
Emat,Kfac,{numer,3}] ]; *)
neldof=Length[Ke];
For [i=1, i<=neldof, i++, ii=eftab[[i]];
For [j=1, j<=neldof, j++, jj=eftab[[j]];
f[[ii]]+=Ke[[i,j]]*u[[jj]]
];
];
]; Return[f];
];

## Figure 12.7. Module to recover node forces including reactions.

srr=szz=srz=stt=Table[0,{numnod}];
Do[{srr[[n]],szz[[n]],stt[[n]],srz[[n]]}=NodeStresses[[n]],{n,1,numnod}];
srrmax=szzmax=sttmax=srzmax=0;
Do[srrmax=Max[Abs[srr[[n]]],srrmax];
szzmax=Max[Abs[szz[[n]]],szzmax];
sttmax=Max[Abs[stt[[n]]],sttmax];
srzmax=Max[Abs[srz[[n]]],srzmax],{n,1,numnod}];
ContourPlotNodeFuncOver2DMesh[NodeCoordinates,ElemNodeLists,
srr,srrmax,Nsub,aspect,"Nodal stress sig-rr"];
ContourPlotNodeFuncOver2DMesh[NodeCoordinates,ElemNodeLists,
szz,szzmax,Nsub,aspect,"Nodal stress sig-zz"];
ContourPlotNodeFuncOver2DMesh[NodeCoordinates,ElemNodeLists,
stt,sttmax,Nsub,aspect,"Nodal stress sig-\[Theta]\[Theta]"];
ContourPlotNodeFuncOver2DMesh[NodeCoordinates,ElemNodeLists,

12–12
12–13 §12.4 POSTPROCESSING

RingNodalStresses[nodcoor_,eletyp_,elenod_,elemat_,
elefab_,eleopt_,u_]:= Module[{numele=Length[elenod],
numnod=Length[nodcoor],numer,type,ne,eNL,eftab,i,ii,j,k,n,
Em,ν,Emat,Kfac,ncoor,ue,ncount,esig,sige,sig,
modname="RingNodalStresses: "},
ncount=Table[0,{numnod}];
sige= Table[0,{numele}]; sig=Table[{0,0,0,0},{numnod}];
{numer,rule,Kfac}=eleopt;
For [ne=1,ne<=numele,ne++,
{type}=eletyp[[ne]];
eNL=elenod[[ne]]; n=Length[eNL]; ue=Table[0,{2*n}];
eftab=Flatten[Table[{2*eNL[[i]]-1,2*eNL[[i]]},{i,1,n}]];
ncoor=Table[nodcoor[[eNL[[i]]]],{i,n}];
Do [ii=eftab[[i]];ue[[i]]=u[[ii]],{i,1,2*n}];
{Em,ν}=elemat[[ne]];
Emat=Em/((1+ν)*(1-2*ν))*{{1-ν,ν,ν,0},{ν,1-ν,ν,0},
{ν,ν,1-ν,0},{0,0,0,1/2-ν}};
(* If [type=="Trig3", esig=Trig3IsoPRingStresses[ncoor,
Emat,Kfac,{numer},ue] ];*)
Emat,Kfac,{numer,"G"},ue] ];
(* If [type=="Trig6", esig=Trig6IsoPRingStresses[ncoor,
Emat,Kfac,{numer,"G"},ue] ];
Emat,Kfac,{numer,"G"},ue] ];
If [type=="Trig10",esig=Trig10IsoPRingStresses[ncoor,
Emat,Kfac,{numer,"G"},ue] ]; *)
esig=Chop[esig]; sige[[ne]]=esig;
For [i=1,i<=n,i++,k=eNL[[i]]; ncount[[k]]++;
Do[ sig[[k,j]]+=esig[[i,j]],{j,4}] ];
];
For [n=1,n<=numnod,n++,k=ncount[[n]];If [k>1,sig[[n]]/=k]
];
Return[sig];
];

## Figure 12.8. Module to compute averaged nodal stresses.

Print["Computed displacements:"];
PrintRingNodeDisplacements[NodeDisplacements];
Print["Node forces including reactions:"];
PrintRingNodeForces[NodeForces];
Print["Averaged nodal stresses:"];
PrintRingNodeStresses[NodeStresses];

## srz,srzmax,Nsub,aspect,"Nodal stress sig-yz"];

12–13
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–14

2b
z
2a

TUBE CROSS
SECTION
p

z
internal pressure p

"SALAMI SLICE"
2 4 6 8 10
h (1) (2) (3) (4) r
r-forces on nodes 1 3 5 7 9
1 and 2 are Kfac pah/2 r=a
r=b

along the z axis

## §12.5. EXAMPLE 1: INTERNALLY PRESSURIZE THICK TUBE

The ﬁrst benchmark problem involves a hollow circular tube of inner radius a and outer radius b
subjected to internal pressure p. The tube extends indeﬁnitely along the z axis. The material is
isotropic with elastic modulus E and Poisson’s ratio ν. See Figure 12.10(a,b).
A “slice” of thickness h is extracted and meshed as shown in Figure 12.10(c), using Nr 4-node quad
ring elements along r and only one along z. Each node must move in the radial direction only. The
two loads applied at the inner nodes are K f ac 12 pah, where K f ac is the subtended angle of the ring
element (typically 2π or 1).
The exact stress distribution2 is
a2 a2 2a 2 ν
σrr = p (1 − b2 /r 2 ), σθ θ = p (1 + b2 /r 2 ), σzz = p , τr z = 0.
b2 − a 2 b2 − a 2 b2 − a 2
(12.9)
Transforming this stress ﬁeld to strains through the inverse of E gives the exact strain ﬁeld, which
veriﬁes ezz = 0. The radial displacement is recovered from u r = r eθ θ and is

a 2 (1 + ν)(b2 + r 2 (1 − 2ν))
ur = p . (12.10)
(b2 − a 2 )Er

2 Taken from S. Timoshenko and J. N. Goodier, Theory of Elasticity, McGraw-Hill, 2nd ed., 1951, Chapter 4. The solution
is for a condition of plane strain.

12–14
12–15 §12.5 EXAMPLE 1: INTERNALLY PRESSURIZE THICK TUBE

ClearAll[Em,ν,a,b,h,p,Kfac,Ner,Nez];
Em=1000.; ν=0; Ner=4; Nez=1;
Kfac=1; a=4; b=10; h=1; p=10.0; aspect=h/(b-a);
(* Define FEM model *)
MeshCorners=N[{{a,0},{b,0},{b,h},{a,h}}];
NodeCoordinates=GenerateRingNodeCoordinates[MeshCorners,Ner,Nez];
numnod=Length[NodeCoordinates]; numele=Length[ElemNodeLists];
ElemMaterial= Table[{Em,ν},{numele}];
ElemFabrication= {}; ElemBodyForces= {};
FreedomValues= Table[{0,0},{numnod}];
FreedomValues[[1]]=FreedomValues[[2]]={Kfac*p*a*h/2,0};
FreedomTags=Table[{0,1},{numnod}];
ElemOptions={True,2,Kfac};
Plot2DElementsAndNodes[NodeCoordinates,ElemNodeLists,aspect,
"Thick ring mesh",True,True];
(* Solve problem and print results *)
{NodeDisplacements,NodeForces,NodeStresses}=RingSolution[
NodeCoordinates,ElemTypes,ElemNodeLists,
ElemMaterial,ElemFabrication,ElemBodyForces,
ElemOptions,FreedomTags,FreedomValues];
Print["Computed displacements:"];
PrintRingNodeDisplacements[NodeDisplacements];
Print["Node forces including reactions:"];
PrintRingNodeForces[NodeForces];
Print["Averaged nodal stresses:"];
PrintRingNodeStresses[NodeStresses];

Figure 12.11. Thick tube under internal pressure: script for running the FEM analysis.

2 4 6 8 10

1 2 3 4
1 3 5 7 9

## Node ur uz node sigrr sigzz sigθθ sigrz

1 0.0546355 0 1 −6.60399 0 13.3193 0
2 0.0546355 0 2 −6.60399 0 13.3193 0
3 0.0447295 0 3 −4.79537 0 7.95551 0
4 0.0447295 0 4 −4.79537 0 7.95551 0
5 0.0402494 0 5 −2.12928 0 5.68578 0
6 0.0402494 0 6 −2.12928 0 5.68578 0
7 0.0383417 0 7 −0.79839 0 4.48206 0
8 0.0383417 0 8 −0.79839 0 4.48206 0
9 0.0378542 0 9 −0.32498 0 3.76736 0
10 0.0378542 0 10 −0.32498 0 3.76736 0

Figure 12.12. Thick tube under internal pressure: computed displacements and stress results.

The numerical results that follow were computed for a = 4, b = 10, h = 2, p = 10, E = 1000
and ν = 0. Figure 12.11 shows the script for running the FEM analysis with 4 and 1 elements
in the radial and longitudinal directions, respectively. Figure 12.12 show the script for graphical
comparison of the FEM results and the exact analytical solution.
The displacements and stress results of the analysis are shown in Figure 12.13. The comparison
with the exact solution is shown in the plots of ﬁgure 12.14. It is seen there that the compted
displacements and the hoop stress are satisfactory captured with this extremely coarse mesh. The

12–15
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–16

urexact[r_]:=p*a^2*(1+ν)*(b^2+r^2*(1-2*ν))/((b^2-a^2)*Em*r);
sigrrexact[r_]:=p*a^2/(b^2-a^2)*(1-b^2/r^2);
sigθθ exact[r_]:=p*a^2/(b^2-a^2)*(1+b^2/r^2);
urlist=Table[{NodeCoordinates[[2*n-1,1]],
NodeDisplacements[[2*n-1,1]]},{n,Ner+1}];
pFEM=ListPlot[urlist,PlotJoined->True,DisplayFunction->Identity];
pexact=Plot[urexact[r],{r,a,b},DisplayFunction->Identity];
Show[Graphics[AbsoluteThickness[2]],pexact,Graphics[RGBColor[1,0,0]],
pFEM,DisplayFunction->\$DisplayFunction,
TextStyle->{FontFamily->"Times",FontSize->12},
GridLines->Automatic,Axes->True,PlotRange->All,
sigrrlist=Table[{NodeCoordinates[[2*n-1,1]],
NodeStresses[[2*n-1,1]]},{n,Ner+1}];
pFEM=ListPlot[sigrrlist,PlotJoined->True,DisplayFunction->Identity];
pexact=Plot[sigrrexact[r],{r,a,b},DisplayFunction->Identity];
Show[Graphics[AbsoluteThickness[2]],pexact,Graphics[RGBColor[1,0,0]],
pFEM,DisplayFunction->\$DisplayFunction,
TextStyle->{FontFamily->"Times",FontSize->12},
GridLines->Automatic,Axes->True,PlotRange->All,
sigθθ list=Table[{NodeCoordinates[[2*n-1,1]],
NodeStresses[[2*n-1,3]]},{n,Ner+1}];
pFEM=ListPlot[sigθθ list,PlotJoined->True,DisplayFunction->Identity];
pexact=Plot[sigθθ exact[r],{r,a,b},DisplayFunction->Identity];
Show[Graphics[AbsoluteThickness[2]],pexact,Graphics[RGBColor[1,0,0]],
pFEM,DisplayFunction->\$DisplayFunction,
TextStyle->{FontFamily->"Times",FontSize->12},
GridLines->Automatic,Axes->True,PlotRange->All,
PlotLabel->"Hoop stress sigθθ (black=exact,red=FEM)"];

Figure 12.12. Thick tube under internal pressure: script for comparing FEM and exact solution.

Hoop stress sigqqHblack=exact,red=FEML
0.055 14
5 6 7 8 9 10
0.0525
-2 12
0.05
0.0475 -4 10

0.045 -6 8
0.0425
-8 6
5 6 7 8 9 10
0.0375 -10
5 6 7 8 9 10

Figure 12.14. Thick tube under internal pressure: comparison of FEM with exact solution.

radial stress is correctly matched inside the tube, but the value on the hole edge, which should be
σrr = − p, is not well predicted. To get this stress within engineering accuracy, 16 elements are
required.
If Poisson’s ratio is increased, the FEM results gradually lose accuracy if the number of elements
is kept the same. As ν approaches 12 , (for example, ν = 0.49) the deterioration in stress predictions
accelerates, and absurd results are predicted on the edges. A ﬁner mesh cures the problem inside
the tube but signiﬁcant loss of accuracy still occurs at r = a.

12–16
12–17 §12.6 EXAMPLE 2: ROTATING THIN DISK

2b
2a

z
DISK CROSS
SECTION
ω

FEM DISCRETIZATION
2 4 6 8 10
h (1) (2) (3) (4) r
1 3 5 7 9
r=a
r=b

## §12.6. EXAMPLE 2: ROTATING THIN DISK

The second benchmark problem is a hollow thin circular disk of thickness h, inner radius a and
outer radius b, which spins about the z axis with constant angular frequency ω. The material is
isotropic with elastic modulus E and Poisson’s ratio ν. See Figure 12.15(a,b).
The FEM discretization is shown in Figure 12.15(c). It uses Nre 4-node quad ring elements along r
(Nre = 4 in the ﬁgure) and only one along z. All nodes are free to move in both directions, except
that one node (node 1) is constrained to move in the z direction to preclude the RBM motion. The
only load is a centrifugal body force br = ρω2r , where ρ is the mass density.
The exact stress distribution3 is

## 3+ν 2 3+ν 2 (1 + 3ν)r 2

σrr = r ω2 (b + a 2 − a 2 b2 /r 2 − r 2 ), σθ θ = r ω2 (b + a 2 + a 2 b2 /r 2 − ,
8 8 (3 + ν)
(12.11)
others zero. Transforming this stress ﬁeld to strains through the inverse of E gives the exact strain
ﬁeld, which veriﬁes ezz = 0. The radial displacement is recovered from u r = r eθ θ and is

## ((a 2 (3 + ν)(−(r 2 (−1 + ν)) + b2 (1 + ν)) − r 2 (−1 + ν)(−(r 2 (1 + ν)) + b2 (3 + ν)))r ω2 )

ur =
8Er
(12.12)

3 Taken from S. Timoshenko, Theory of Elasticity. MacGraw-Hill, 4th ed., Chapter 2 19??. The solution is for a condition
of plane stress.

12–17
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–18

ClearAll[Em,ν,a,b,h,Kfac,ρ,ω,Ner,Nez];
Em=1000.; ν=1/3; Ner=4; Nez=1;
Kfac=1; a=4; b=10; h=1; aspect=h/(b-a); ρ=3.0; ω=0.5;
(* Define FEM model *)
MeshCorners=N[{{a,0},{b,0},{b,h},{a,h}}];
NodeCoordinates=GenerateRingNodeCoordinates[MeshCorners,Ner,Nez];
numnod=Length[NodeCoordinates]; numele=Length[ElemNodeLists];
ElemMaterial= Table[{Em,ν},{numele}];
ElemFabrication= {};
FreedomTags= Table[{0,0},{numnod}]; FreedomTags[[1]]={0,1};
FreedomValues= Table[{0,0},{numnod}];
ElemBodyForces= Table[{0,0},{numele}];
For [ne=1,ne<=numele,ne++, eNL=ElemNodeLists[[ne]];
ncoor=Table[NodeCoordinates[[eNL[[i]]]],{i,4}];
{{r1,z1},{r2,z2},{r3,z3},{r4,z4}}=ncoor;
ElemBodyForces[[ne]]=ρ*ω^2*{{r1,0},{r2,0},{r3,0},{r4,0}} ];
ElemOptions={True,2,Kfac};
Plot2DElementsAndNodes[NodeCoordinates,ElemNodeLists,aspect,
"Rotating disk mesh",True,True];
(* Solve problem and display results *)
{NodeDisplacements,NodeForces,NodeStresses}=RingSolution[
NodeCoordinates,ElemTypes,ElemNodeLists,
ElemMaterial,ElemFabrication,ElemBodyForces,
ElemOptions,FreedomTags,FreedomValues];
Print["Computed displacements:"];
PrintRingNodeDisplacements[NodeDisplacements];
Print["Node forces including reactions:"];
PrintRingNodeForces[NodeForces];
Print["Averaged nodal stresses:"];
PrintRingNodeStresses[NodeStresses];

Figure 12.16. Rotating thin disk: script for running the FEM analysis.

2 4 6 8 10
1 2 3 4
1 3 5 7 9

## Node ur uz node sigrr sigzz sigθθ sigrz

1 0.255711 0 1 12.3626 2.17846 67.5211 −0.781042
2 0.255711 −0.0244494 2 12.3626 2.17846 67.5211 0.781042
3 0.239405 −0.00312417 3 8.04755 −0.384588 45.402 −0.533743
4 0.239405 −0.0213253 4 8.04755 −0.384588 45.402 0.533743
5 0.234836 −0.00426995 5 9.0946 −0.878575 35.9983 −0.405955
6 0.234836 −0.0201795 6 9.0946 −0.878575 35.9983 0.405955
7 0.231569 −0.00637181 7 6.15515 0.0637133 29.1534 −0.467814
8 0.231569 −0.0180776 8 6.15515 0.0637133 29.1534 0.467814
9 0.224084 −0.00801246 9 2.91302 0.336958 23.3714 −0.410162
10 0.224084 −0.016437 10 2.91302 0.336958 23.3714 0.410162

## Figure 12.17. Rotating thin disk: stress-displacement analysis results.

The numerical results that follow were computed for a = 4, b = 10, h = 1, E = 1000, ν = 1/3,
ρ = 3 and ω = 0.5,

12–18
12–19 §12.6 EXAMPLE 2: ROTATING THIN DISK

urexact[r_]:=((a^2*(3+ν)*(-(r^2*(-1+ν))+b^2*(1+ν))- r^2*(-1+ν)*
(-(r^2*(1+ν))+b^2*(3+ν)))*ρ*ω^2)/(8*Em*r);
sigrrexact[r_]:=ρ*ω^2*(3+ν)/8*(b^2+a^2-a^2*b^2/r^2-r^2);
sigθθ exact[r_]:=ρ*ω^2*(3+ν)/8*(b^2+a^2+a^2*b^2/r^2-(1+3*ν)*r^2/(3+ν));
urlist=Table[{NodeCoordinates[[2*n-1,1]],
NodeDisplacements[[2*n-1,1]]},{n,Ner+1}];
pFEM=ListPlot[urlist,PlotJoined->True,DisplayFunction->Identity];
pexact=Plot[urexact[r],{r,a,b},DisplayFunction->Identity];
Show[Graphics[AbsoluteThickness[2]],pexact,Graphics[RGBColor[1,0,0]],
pFEM,DisplayFunction->\$DisplayFunction,
TextStyle->{FontFamily->"Times",FontSize->12},
GridLines->Automatic,Axes->True,PlotRange->All,
sigrrlist=Table[{NodeCoordinates[[2*n-1,1]],
NodeStresses[[2*n-1,1]]},{n,Ner+1}];
pFEM=ListPlot[sigrrlist,PlotJoined->True,DisplayFunction->Identity];
pexact=Plot[sigrrexact[r],{r,a,b},DisplayFunction->Identity];
Show[Graphics[AbsoluteThickness[2]],pexact,Graphics[RGBColor[1,0,0]],
pFEM,DisplayFunction->\$DisplayFunction,
TextStyle->{FontFamily->"Times",FontSize->12},
GridLines->Automatic,Axes->True,PlotRange->All,
sigθθ list=Table[{NodeCoordinates[[2*n-1,1]],
NodeStresses[[2*n-1,3]]},{n,Ner+1}];
pFEM=ListPlot[sigθθ list,PlotJoined->True,DisplayFunction->Identity];
pexact=Plot[sigθθ exact[r],{r,a,b},DisplayFunction->Identity];
Show[Graphics[AbsoluteThickness[2]],pexact,Graphics[RGBColor[1,0,0]],
pFEM,DisplayFunction->\$DisplayFunction,
TextStyle->{FontFamily->"Times",FontSize->12},
GridLines->Automatic,Axes->True,PlotRange->All,
PlotLabel->"Hoop stress sigθθ (black=exact,red=FEM)"];

Figure 12.18. Rotating thin disk: script for comparing FEM results with the analytical solution.

12
0.255
0.25 10 60
0.245 8
50
0.24 6
0.235 4 40
2
5 6 7 8 9 10
0.225
5 6 7 8 9 10
5 6 7 8 9 10

Figure 12.18. Rotating thin disk: comparison of FEM results with exact solution for u r , σrr and σθθ .

Figure 12.16 shows the script for running the FEM analysis with 4 and 1 elements in the radial and
longitudinal directions, respectively. The displacements and stress printed results of the are shown
in Figure 12.17.
Figure 12.18 show the script for graphical comparison of the FEM results and the exact analytical
solution. The comparisons with u r , σrr and σθ θ are shown in the plots of ﬁgure 12.19. It is seen there
that the radial displacement and hoop stress variations are satisfactory captured with this extremely
coarse mesh. The radial stress is poorly predicted, specially at the hole edge, which should be
σrr = 0, To get this stress within engineering accuracy, at least 16 elements are required.

12–19
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–20

2R

z
Plan View

z
Simple Support

## ;;;; FEM DISCRETIZATION

2 4 6 8 10 12 14 16 18
h (1) (2) (3) (4) (1) (2) (3) (4) 17 r

;;
1 3 5 7 9 11 13 15

r=R

## Figure 12.20. Centrally loaded SS circular plate.

From Figure 12.17 it can be observed that the model gives nonzero σzz and τr z stresses, which
should be zero in a plane stress model. If ν = 0 the values are much closer to zero.
As in the previous benchmark, as ν approaches 12 , (for example, ν = 0.49) the deterioration in
stress predictions accelerates, and absurd results are predicted. A ﬁner mesh cures the problem
inside the disk but signiﬁcant loss of accuracy still occurs at the disk free edges.

## §12.7. EXAMPLE 3: CENTRALLY LOADED SS CIRCULAR PLATE

The third benchmark problem is a centrally loaded simply-supported circular plate. the plate has
radius R and thickness h. The point load of magnitude P acts downward at the plate center. The
material is isotropic with elastic modulus E and Poisson’s ratio ν. See Figure 12.20(a,b).
The FEM discretization is shown in Figure 12.20(c). It uses Nre 4-node quad ring elements along
r (Nre = 8 in the ﬁgure) and only one along z. The SS condition is speciﬁed by a roller at the outre
edge node (the point should not be ﬁxed, as it should be able to move in the radial direction). The
two nodes at the center r = 0 are constrained against radial motion. The load is placed on the upper
node at r = 0 and received the value P K f ac /(2π ).
The exact stress distribution4 for a thin (Kirchhoff) plate has the bending moments
 
P R P R
Mrr = (1 + ν) log , Mθ θ = (1 + ν) log + 1 − ν . (12.13)
4π r 4π r

4 Taken from S. Timoshenko and S. Woinowsky-Krieger, Theory of Plates and Shells, McGraw-Hill, 2nd ed., 1959,
Chapter 2.

12–20
12–21 §12.7 EXAMPLE 3: CENTRALLY LOADED SS CIRCULAR PLATE

ClearAll[Em,ν,R,h,p,Kfac,Ner,Nez,r,z];
Em=1000.; ν=0; Ner=8; Nez=1;
Kfac=2*Pi; R=10; h=1; aspect=h/R; P=10;
(* Define FEM model *)
MeshCorners=N[{{0,-h/2},{R,-h/2},{R,h/2},{0,h/2}}];
NodeCoordinates=GenerateRingNodeCoordinates[MeshCorners,Ner,Nez];
numnod=Length[NodeCoordinates]; numele=Length[ElemNodeLists];
ElemMaterial= Table[{Em,ν},{numele}];
ElemFabrication= {}; ElemBodyForces= {};
FreedomTags= Table[{0,0},{numnod}];
FreedomTags[[1]]=FreedomTags[[2]]={1,0};
FreedomTags[[numnod-1]]={0,1};
FreedomValues= Table[{0,0},{numnod}];
FreedomValues[[2]]={0,N[-P*Kfac/(2*Pi)]};
ElemOptions={True,2,Kfac};
Plot2DElementsAndNodes[NodeCoordinates,ElemNodeLists,aspect,
"Thin SS circular plate",True,True];
(* Solve problem and display results *)
{NodeDisplacements,NodeForces,NodeStresses}=RingSolution[
NodeCoordinates,ElemTypes,ElemNodeLists,
ElemMaterial,ElemFabrication,ElemBodyForces,
ElemOptions,FreedomTags,FreedomValues];
Print["Computed displacements:"];
PrintRingNodeDisplacements[NodeDisplacements];
Print["Node forces including reactions:"];
PrintRingNodeForces[NodeForces];
Print["Averaged nodal stresses:"];
PrintRingNodeStresses[NodeStresses];

Figure 12.21. Centrally loaded SS circular plate: script for running the FEM analysis.

The stresses at the upper and lower surfaces are σrr = ±12Mrr (h/2)/ h 3 = ±6Mrr / h 2 and
σθθ = ±12Mθ θ (h/2)/ h 3 = ±6Mθ θ / h 2 The transverse displacement is
 
P 3+ν 2 r
uz = − (a − r ) + 2r log
2 2
. (12.14)
16π D 1 + ν R

## where D = Eh 3 /(12(1 − ν 2 )) is the plate rigidity.

The numerical results that follow were computed for R = 10, h = 1, E = 1000, ν = 1/3, and
P = 10. Figure 12.21 shows the script for running the FEM analysis with 8 and 1 elements in the
radial and longitudinal directions, respectively. The displacements and stress printed results of the
are shown in Figure 12.22.
Figure 12.23 show the script for graphical comparison of the FEM results and the exact analytical
solution. The comparisons with u z , σrr and σθ θ are shown in the plots of ﬁgure 12.24. It is seen
there that the 8-element mesh does a good job with all quantities except for radial stress near the
center. The discrepancy there is inevitable becuase of the concentration effects of the point load.
From Figure 12.17 it can be observed that the model gives nonzero σzz and τr z stresses, which
should be zero in a Kirchhoff plate bending model. If ν = 0 the values are much closer to zero.

12–21
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–22

## Thin SS circular plate

2 4 6 8 10 12 14 16 18
1 2 3 4 5 6 7 8
1 3 5 7 9 11 13 15 17

## Node ur uz node sigrr sigzz sigθθ sigrz

1 0 −0.668684 1 16.0918 −5.68759 16.0918 14.7031
2 0 −0.674372 2 −16.0918 −5.68759 −16.0918 17.2096
3 0.0201148 −0.631926 3 11.2016 0.578633 15.7157 −0.02329
4 −0.0201148 −0.631348 4 −11.2016 0.578633 −15.7157 1.09662
5 0.0280041 −0.568227 5 5.20148 −0.0880224 10.9175 0.346082
6 −0.0280041 −0.568315 6 −5.20148 −0.0880224 −10.9175 0.23335
7 0.0331185 −0.490176 7 3.44022 0.0149693 8.72616 0.147734
8 −0.0331185 −0.490161 8 −3.44022 0.0149693 −8.72616 0.164716
9 0.0366047 −0.401896 9 2.3363 −0.00311393 7.26742 0.10501
10 −0.0366047 −0.401899 10 −2.3363 −0.00311393 −7.26742 0.102535
11 0.0389593 −0.306627 11 1.54134 0.00259233 6.2016 0.0770656
12 −0.0389593 −0.306625 12 −1.54134 0.00259233 −6.2016 0.075774
13 0.040458 −0.206714 13 0.926306 −0.00957223 5.37344 0.055775
14 −0.040458 −0.206724 14 −0.926306 −0.00957223 −5.37344 0.0639061
15 0.041275 −0.104058 15 0.428532 0.0432483 4.70242 0.0677792
16 −0.041275 −0.104015 16 −0.428532 0.0432483 −4.70242 0.0298345
17 0.0415293 0 17 0.203464 −0.199296 4.14122 0.0939349
18 −0.0415293 −0.000199296 18 −0.203464 −0.199296 −4.14122 −0.00308274

## Figure 12.22. Centrally loaded SS circular plate: stress-displacement analysis results.

12–22
12–23 §12.7 EXAMPLE 3: CENTRALLY LOADED SS CIRCULAR PLATE

Dp=Em*h^3/(12*(1-ν^2)); eps=0.01;
uzexact[r_]:=-P/(16*Pi*Dp)*((3+ν)/(1+ν)*(R^2-r^2)+2*r^2*Log[r/R]);
Print["exact center deflection: ", -P*R^2/(16*Pi*Dp)*(3+ν)/(1+ν)];
Mrrexact[r_]:=P/(4*Pi)*(1+ν)*Log[R/r];
Mθθ exact[r_]:=P/(4*Pi)*((1+ν)*Log[R/r]+1-ν);
sigrrexact[r_,z_]:=-12*Mrrexact[r]*z/h^3;
sigθθ exact[r_,z_]:=-12*Mθθ exact[r]*z/h^3;
uzlist=Table[{NodeCoordinates[[2*n-1,1]],
NodeDisplacements[[2*n-1,2]]},{n,Ner+1}];
pFEM=ListPlot[uzlist,PlotJoined->True,DisplayFunction->Identity];
pexact=Plot[uzexact[r],{r,eps*R,R},DisplayFunction->Identity];
Show[Graphics[AbsoluteThickness[2]],pexact,Graphics[RGBColor[1,0,0]],
pFEM,DisplayFunction->\$DisplayFunction,
TextStyle->{FontFamily->"Times",FontSize->12},
GridLines->Automatic,Axes->True,PlotRange->All,
PlotLabel->"Transverse (z) displacements (black=exact,red=FEM)"];
sigrrlist=Table[{NodeCoordinates[[2*n-1,1]],
NodeStresses[[2*n-1,1]]},{n,Ner+1}];
pFEM=ListPlot[sigrrlist,PlotJoined->True,DisplayFunction->Identity];
pexact=Plot[sigrrexact[r,-h/2],{r,eps*R,R},DisplayFunction->Identity];
Show[Graphics[AbsoluteThickness[2]],pexact,Graphics[RGBColor[1,0,0]],
pFEM,DisplayFunction->\$DisplayFunction,
TextStyle->{FontFamily->"Times",FontSize->12},
GridLines->Automatic,Axes->True,PlotRange->All,
sigθθ list=Table[{NodeCoordinates[[2*n-1,1]],
NodeStresses[[2*n-1,3]]},{n,Ner+1}];
pFEM=ListPlot[sigθθ list,PlotJoined->True,DisplayFunction->Identity];
pexact=Plot[sigθθ exact[r,-h/2],{r,eps*R,R},DisplayFunction->Identity];
Show[Graphics[AbsoluteThickness[2]],pexact,Graphics[RGBColor[1,0,0]],
pFEM,DisplayFunction->\$DisplayFunction,
TextStyle->{FontFamily->"Times",FontSize->12},
GridLines->Automatic,Axes->True,PlotRange->All,
PlotLabel->"Hoop stress sigθθ (black=exact,red=FEM)"];

Figure 12.23. Centrally loaded SS circular plate: script for comparing FEM results with the analytical solution.

Transverse HzLdisplacements Hblack=exact,red=FEM) Radial stress sigrr Hblack=exact,red=FEML Hoop stress sigqqHblack=exact,red=FEML

2 4 6 8 10 20 25
-0.1
-0.2 20
15
-0.3
-0.4 10 15
-0.5
-0.6 5 10
-0.7
2 4 6 8 10
2 4 6 8 10

Figure 12.24. Centrally loaded SS circular plate: comparison of FEM results with exact solution for u r , σrr and σθθ .

12–23
Chapter 12: A COMPLETE AXISYMMETRIC FEM PROGRAM 12–24

## Homework Exercises for Chapter 12

A Complete Axisymmetric FEM Program

EXERCISE 12.1
[C:15] This exercise deals with the thick-tube benchmark example discussed in §12.5. The script for this
exercise is in Cells 15A and 15B of the Notebook Quad4.SOR.nb supplied with this Chapter. The cells should
be run sequentially.
(a) Repeat the 4-element analysis using ν = 0.45 and ν = 0.499. Describe what happens to the accuracy of
displacements and stresses when compared with the exact solution.
(b) Can a more reﬁned mesh ﬁx the problems noted in (a)? To check, run 16 elements along the radial
direction and report if things have improved.

EXERCISE 12.2
[C:20] A concrete pile embedded in a soil half-space, as deﬁned in Figure E12.1. This problem is solved using a
very coarse mesh in Cell 18 of the Notebook Quad4.SOR.nb supplied with this Chapter. Repeat the analysis us-
ing a more reﬁned mesh, like that suggested in the ﬁgure. Module GenerateGradedRingNodeCoordinates
may be used to generate a graded regular mesh.
The total force applied to pile is P = 500000 lbs (this may be assumed to be uniformly distributed on the
top pile surface, or placed as a point load). Other data: pile modulus E P = 300, 000 psi, soil modulus
E S = E P /50, Poisson’s ratio for pile ν P = 0.1, Poisson’s ratio for soil ν S = 0.40, pile diameter d = 10 in,
pile length L = 250 in, and mesh z-length H = 1.2L. Truncate the mesh at R = 80 in. Neglect all body
forces.
Nodes on the “soil truncation boundary” should be ﬁxed. Points on the z axis r = 0 should be on vertical
rollers except for 1.
The most interesting numerical results are: (i) the top and bottom vertical displacement of the pile, (ii) the
reaction forces at the bottom of the pile, and (iii) the normal stress σzz in the pile.

EXERCISE 12.3
[C:20] The spinning Mother Earth. Model one quadrant of the planet cross section with an axisymmetric ﬁnite
element mesh as sketched in Figure E19.3 (note that all elements are 4-node quadrilaterals). The problem is
solved in Cell 19 of the Notebook Quad4.SOR.nb supplied with this Chapter, using a very coarse mesh of
only 4 elements.

Use the Kg-force/m/sec unit system for this problem. The Earth spins with angular velocity ω = 2π rad/24hrs
= (2π/86400) sec−1 about the z axis. The planet radius is R = 6370 Km = 6.37 106 m. For E take 1/3 of the
rigidity of steel, or E = 7 105 Kg/cm2 = 7 109 Kg/m2 as Love (Theory of Elasticity) recommends; Poisson’s
ratio ν = 0.3 (this is my own guess), and mass density ρ = 5.52 times the water density. [Watch out for units:
the centrifugal body force ρω2 r should come up in Kg/m3 .] All gravitational ﬁeld effects (self weight) are
ignored.
(a) Get the equatorial “bulge” and the polar “ﬂatening” in Km, and the maximum stress in MPa.
(b) Where do the maximum stresses occur?

12–24
12–25 Exercises

;
z
R
d/2

; ;
(b) 9 18 27 36 45 r
z

(a)

;; ;
CONCRETE
P 8 44
PILE
r

;
7 43

; ;
L
L H 6 42
SOIL

;
5 41

AXISYMMETRIC
FEM MODEL
d 4 40
3 39

2 38

1 10 19 28 37

12–25