Plane, axisymmetric and 3D models

in
Mentat & MARC

Tutorial with Background and Exercises

Eindhoven University of Technology

Department of Mechanical Engineering
Piet Schreurs January 28, 2010
Contents
1 Finite element method 3

2 Plane stress and plane strain 6

2.1 Background : Theory and element formulation . . . . . . . . . . . . . . . . . . . . 6
2.2 Plate with a central hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Modelling and analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.3 Dynamic boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Axle bearing with radial load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.1 Modelling and analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.3 Quadratic elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Orthotropic plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.5.1 Large thin plate with a cirular hole . . . . . . . . . . . . . . . . . . . . . . . 25
2.5.2 Plate with a square hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.5.3 Thick-walled pressurized cylinder . . . . . . . . . . . . . . . . . . . . . . . . 29
2.5.4 Rotating engineering parts . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3 Axisymmetry 34
3.1 Background : Axisymmetric modelling . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2 Axle bearing with axial load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.1 Modelling, analysis and results . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.3.1 Circular plate with a circular hole . . . . . . . . . . . . . . . . . . . . . . . 40
3.3.2 Rotating engineering parts . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

4 Three-dimensional models 43
4.1 Background : Three-dimensional elements . . . . . . . . . . . . . . . . . . . . . . . 43
4.2 Three-dimensional model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.3 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.3.1 Inhomogeneous plate with a central circular hole . . . . . . . . . . . . . . . 46

2
1 Finite element method
In this section it is described how an approximate solution for the equilibrium equation can be
determined with the finite element method. This description is very concise and detailed deriva-
tions are not included.

Weighted residual formulation

The equilibrium equation reads :

~ + ~
q = ~0 ~x V

Analytical solutions for this partial differential equation only can be found for very simple prob-
lems. For the more realistic cases only approximate solutions can be determined, using a numerical
technique. The finite element method is widely used for this purpose. For the application of this
method the differential equation is transformed into an integral equation. This transformation is
formulated as follows :

The requirement that in each material point ~x V of the body the equi-
librium equation

~ + ~
q = ~0

must be satisfied, is equivalent to the requirement that for every allowed

function w(~
~ x) the integral over the volume V
Z h i
w~ ~ + ~
q dV = 0
V

is satisfied.

The function w(~

~ x) is the weighting function and the integral the weighted residual formulation of
the equilibrium equation. These names become obvious, when it is realized that for an approximate
solution, the equilibrium equation is not satisfied exactly and that an error ~ + ~
q 6= ~0 will
remain. This local error is smeared out over the whole volume V of the body by the integral,
employing a weighting function w(~
~ x).
The weighted residual integral is the starting point for application of the finite element method
and will now be further elaborated.

Weak formulation

To relax the continuity requirements concerning the approximate stress field (~x), the derivative
of the stress tensor is removed from the integral. This can be done by simple partial integration
of the divergence term.

~ ( w)
~ = ( ~ w)
~ : T + w ~ )
~ ( substitution
Z h i
~ ( w)
~ (~ w)
~ : T + w q dV = 0
~ ~ w
~
V
Z Z Z
~ w)
( ~ T : dV = w q dV +
~ ~ ~n w
~ dS w
~
V V S

Applying Gauss theorem transformed the first term into an integral over the boundary S of V ,
where ~n is the unity vector in a boundary point.

3
The stress tensor can be related to the strain tensor , which can be expressed in the gradient of
the displacement ~u of the material points. For linear elastic material behavior, we have

~ u)
= 4 C : = 4 C : (~

where 4 C is the fourth-order material tensor. In a boundary point we have ~n = ~p, where p~ is
the stress vector in that point. The weighted residual integral becomes :
Z Z Z
~ w)
( ~ u) dV = w
~ T : 4 C : (~ q dV + w
~ ~ ~ p~ dS w ~
V V S

Discretisation

The integrals over V and S are written as a sum of integrals over a number of sub-volumes
and -boundaries, in which V and S are subdivided. These volumes are the elements and their
boundaries as far as these are located on the external boundary S of V , with volume V e and
area S e , respectively. The discretised weighted residual integral becomes :
XZ XZ XZ
(~ w) ~ u) dV e =
~ T : 4 C : (~ w q dV e +
~ ~ ~ p~ dS e
w w~
e Ve e Ve eS S e

Interpolation and integration

The next step in deriving the finite element formulation is the interpolation of the displacement and
the weigthing function. This is done in each element, where a local element coordinate system
with coordinates is used. For three-dimensional elements there are three local coordinates :

T = {1 , 2 , 3 }. Each of these so-called isoparametric coordinates has values between -1 and +1.
The value of a variable ~u and w ~ in an element point with coordinates is expressed in its

values in a number of discrete points of the element, the element nodes. The interpolation function
associated with node i is N i ( ). For ~u and w ~ and for their gradients, we can now write :

T
~ T ~ue

~u = N ( )~u + N ( )~u + N 3 ( )~u3 + = N T ( )~ue ~
1 1 2 2 ~ u = N ~ ~ue = B

T
~Tw

e ~w ~ e
w 1
~ = N ( )w 1 2
~ + N ( )w 2 3
~ + N ( )w 3 T
~ + = N ( )w
~ ~ = N ~ =B
w ~e

The columns ~ue and w ~ e contain the nodal values of ~u and w. ~ The displacement and the weighting

function are interpolated in the same way, which is according to Galerkins method. Substitution of
the interpolations in the contribution of one single element to the total weighted residual integral,
gives :
Z Z Z
~Tw ~ T ~ue ) dV e = w
T T
(B ~ e )T : 4 C : (B ~ e N ~
q dV e + w ~ e N ~p dS e ~e
w

Ve Ve Se

Z Z Z
~ T dV e ~ue = w
T T T
w~e B ~ 4C B ~ e N ~ q dV e + w~ e N p~ dS e ~e
w

Ve Ve Se
eT eT e
w
~ e e
K ~u = w
~ f~e w
~ e

e
where K e is the element stiffness matrix and f~e the column with the external nodal forces. Cal-

culating the element stiffness matrix implies integration over the element volume. The integral,
whose integrand is a function of the local coordinates , can not be calculated analytically. It is
done numerically by evaluating the integrand in a number of discrete so-called integration points

4
and adding the results after multiplication by a weighting factor. For a three-dimensional element
we can write this as

Z1 Z1 Z1 X
f (1 , 2 , 3 ) d1 d2 d3 = ci f (1i , 2i , 3i )
1 =1 2 =1 3 =1 i

Assembling

During the assembling procedure the contribution of each element is added to the total weighted
residual integral. The components of the element columns w ~ e and ~ue are inserted at the proper
places in the system columns w ~ and ~u in all nodes. The element
~ and ~u, containing the values of w
e
nodal force f e and the element stiffness matrix K are than added to the system column f~e and
~ e

the system stiffness matrix K, which results in :

w ~ T f~
~ T K ~u = w w
~
e

Set of equations

Because the above equation has to be satisfied for all possible weighting functions w
~ , the unknown
components of ~u and f~e must be determined such that the next set of algebraic equations is
satisfied :

K ~u = f~e

After taking into account the boundary conditions prescribed components of ~u and f~e , the
vectors
unknown displacement components can be solved. It is obvious that all tensors and first
have to be written in components w.r.t. a suitable coordinate system.

Boundary conditions

No external forces are needed for a rigid translation of the structure. Writing the set of equations
in components, this means

k11 k12 k13 a 0
k21 k22 k23 a 0
= or Ka = 0

k31 k32 k33 a 0

.. .. .. .. ..

. . . . . .

The above relation indicates that K is a singular matrix, whos determinant is zero. The set of
equations Ku = f e can never be solved, because the inverse of the singular matrix does not exist.
enough
By prescribing boundary conditions to prevent rigid body motion, the resulting set of
equations can be solved, because the resulting stiffness matrix is no longer singular but regular.

5
2 Plane stress and plane strain
2.1 Background : Theory and element formulation
In the tutorial Truss and beam structures trusses and beams are described and used to model
simple structures with MSC.Marc/Mentat. Although the nodes of the truss and beam elements
can have a displacement (and rotation) in three directions, only one stress is relevant : the axial
stress. For linear elastic material behavior, the stress in a truss is related to the axial strain
= l
l0 , according to = E, where E is the Youngs modulus, which can be measured in a tensile
experiment.

Plane stress

Many structures are build from flat plates, having a thickness, which is much smaller than the
dimensions of the plate in its plane. The figure below shows such a plate, with its plane in the
xy-coordinate plane and with a uniform thickness h0 .

z
y
x

h0

In many cases it is allowed to assume that the plate is loaded in its plane, as shown in the figure.
In that case there will be no bending of the plate and it will stay flat after deformation.

z
y
x
P

In a point P of the plate a column is cut out, with its sides parallel to the xz- and yz-coordinate
planes, respectively, and with dimensions dx dy h0 . This small part of the plate is loaded with
stresses, so forces per area. Because the thickness of the plate is very small, it can be assumed
that these stresses are constant over the thickness. The stress components working on a stress
cube dx dy dz are the normal stresses xx and yy and the shear stresses xy and yx . It can
be shown that xy = yx , so only three stress components are relevant.

6
 3
2
1
P

22
21

11 12

Because there are no stresses working in z-direction, this stress state is referred to as plane stress.
The deformation in point P is described by the strain components xx and yy and the shear xy .
For linear elastic material behavior the next relation holds between stresses and strains :

xx 1 0 xx
E 1
yy = 0 yy
1 2
xy 0 0 1 xy

Due to the stresses xx and yy the thickness of the plate will change. The strain in thickness
direction, so in z-coordinate direction in our case, is :
h
zz = = (xx + yy )
h0 E

Plane strain

The displacement in the z-direction, which is for our plate the direction perpendicular to its plane,
is suppressed. The plate is for instance glued to two parallel rigid bodies, which stay at the same
distance. In that case we have zz = 0, and call the deformation a state of plane strain.
In a state of plane strain, there is generally a stress zz 6= 0. The linear elastic material
behavior is described by :

xx 1 0 xx
yy = E 1 0 yy ; zz = (xx + yy )
(1 + )(1 2)
xy 0 0 1 2 xy

It is noted that problems will occur when 0.5. Stresses will become infinite. This is not
strange, since in a linear elastic material the volume change will be zero for = 0.5. To prevent
numerical problems in MSC.Marc, we use the option CONSTANT DILATATION in the menu GEOMETRY.

Interpolation and integration : the element stiffness matrix

When analyzing plane stress and plane strain problems with the finite element method we have to
use plane stress or plane strain elements. These elements, are in MSC.Marc always located in the
xy-plane. Also the deformation is in the xy-plane and is described by the x- and y-displacement

7
of a number of nodal points, situated on the edges of the elements. The displacement of an
internal point of the element is interpolated between the displacement of the nodes. For this
interpolation, a local coordinate system is used with isoparametric coordinates and , which
have values between -1 and +1. In mathematical terms the interpolation of the displacement u
and v in x- and y-direction, respectively, can be written as follows :
n
X
u(, ) = N 1 (, )u1 + N 2 (, )u2 + N 3 (, )u3 + = N i (, )ui
i=1
n
X
v(, ) = N 1 (, )v 1 + N 2 (, )v 2 + N 3 (, )v 3 + = N i (, )v i
i=1

where N i are the interpolation functions, associated to node i, and ui and v i the nodal displacement
components. The number of element nodes is n.
Calculation of the element stiffness matrix implies integration of a function over the element
volume. This integral can only be evaluated numerically : the integrand is calculated in a discrete
number of internal integration points and these values are added after multiplication with a cer-
tain coefficient. The location of the integration points and the coefficients are fixed for a certain
element type.

Linear and quadratic elements

The figure below shows a 4-node element with all nodes in the four corners of a quadrilateral with
straight sides. Nodal points have a local counterclockwise numbering, which is in accordance with
the MSC.Marc program. The figure also shows the element in the so-called isoparametric space,
 T
where points are identified with the local isoparametric coordinates = .

3
4
4 3

1 2 1 2

Because there are only two nodes on one element side, the displacement of an arbitrary point on
the side can only vary linearly between the two nodal values. A 4-node quadrilateral is therefore
called a linear element. The four interpolation functions, associated with the nodes are :

N 1 = 41 ( 1)( 1) ; N 2 = 14 ( + 1)( 1)
N 3 = 41 ( + 1)( + 1) ; N 4 = 14 ( 1)( + 1)

The element has 4 integration points, which are shown in the figure below. Their local coordinates
are fixed and determined such that the numerical integration is as accurate as possible.

3
4
4
3
1

3 3
2
1
1 2
1

3 3

8
In a quadratic element, the displacement of an arbitrary point is interpolated using interpolation
functions, which are quadratic in the local coordinates and . This element has 8 nodes and is
therefore also referred to as an 8-node element. Four of these nodes are situated in the corners
of the element, the other four are in the middle of the element sides in the undeformed situation.
The numbering of the local nodes is indicated in the figure below and is in accordance with the
MSC.Marc program. The interpolation functions are :

N 1 = 14 ( 1)( 1)( 1) ; N 2 = 41 ( + 1)( 1)( + + 1)

N 3 = 41 ( + 1)( + 1)( + 1) ; N 4 = 41 ( 1)( + 1)( + 1)
N 5 = 12 ( 2 1)( 1) ; N 6 = 21 ( 1)( 2 1)
N 7 = 21 ( 2 1)( 1) ; N 8 = 21 ( 1)( 2 1)

The 8-node element has 9 integration points, the location of which is indicated in the figure.

3
7
4
4 7 3
6
8 6
8

1 5 2 1 5 2

9
7 8
6 1

4 5 15
5

1 2 3
1
5 15

9
2.2 Plate with a central hole
The figure below shows a square plate with a central hole. Relevant dimensions are indicated.

Youngs modulus = 210 [GPa]

Poissons ratio = 0.3 [-]

thickness = 1 [cm]

u [mm]

10 [cm]

16 [cm]

The plate is loaded in its plane. Displacements of left and right edges (x = 8 [cm]) are prescribed :
u = 0.1 [mm]. In y-direction the displacement of the edge points is free. Top and bottom edge
(y = 8 [cm]) are also free. This load leads to a plane stress state in the plate.
The material of the plate is isotropic and can be assumed to be linearly elastic. Youngs
modulus and Poissons ratio are known : E = 210 [GPa] and = 0.3 [-].
Deformation of and stresses in the plate will be determined using MARC and therefore a model
must be made with Mentat. Considering symmetry and load learns that only a quarter of the
plate has to be modelled. We choose the part with 0 (x, y) 8 [cm]. The correct symmetry
conditions must of course be prescribed as boundary conditions.

10
2.2.1 Modelling and analysis

We start in the MAIN MENU and go to MESH GENERATION. We will use the
Cartesian coordinate system. This is shown on the computer screen as a two-
dimensional grid in the xy-plane. In the grid points we can define (nodal) points
and other things by clicking with the left mouse button. In the grid settings
the grid dimensions and the spacing between the grid points can be specified.
The program does not know of dimensions and units, so we have to choose
them and stick to them during the modelling and analysis. Here, dimensions
will be expressed in unit of meters. We choose the grid point spacing to be
0.01 meter.

(MAIN MENU)(PREPROCESSING)
MESH GENERATION
(COORDINATE SYSTEM)
SET
U DOMAIN : -0.1, 0.1
U SPACING :
0.01
V DOMAIN : -0.1, 0.1
V SPACING : 0.01
GRID
FILL
RETURN

In the tutorial for Truss and beam structures we used only CURVES, but this
is not enough any more; we have to do more. We want to define a SURFACE
and subdivide this into elements. A simple method is used here, which can
be applied in many cases. First two CURVES are defined, one POLY LINE and a
quarter of a circle, an ARC. The POINTS which define the CURVES can be located
in grid points by clicking the left mouse button.
After defining the CURVES, a surface of the type RULED is made. The idea is
that a stick is placed with its begin and end point on two separate CURVES
and is subsequently rolled over them, thus describing a SURFACE in space.

CURVE TYPE
POLYLINE
RETURN
(CRVS) ADD

Define three points of the POLYLINE.

Close the window with END LIST.

CURVE TYPE
(ARCS) CENTER/POINT/POINT
RETURN
(CRVS) ADD

Define the required (see Command-screen) points of the ARC.

When the ARC is drawn in the wrong direction, click UNDO and define the latter
two ARC points in swapped order.

11
 SURFACE TYPE
RULED
RETURN
(SRFS) ADD

Click on the two CURVES.

A strange surface may appear because of different orientation of the two

curves. In that case we have to UNDO the last action (definition of the surface)
and flip one of the two curves in the CHECK menu with FLIP CURVE. The
surface must then be redefined. When this is done successfully, the surface is
now going to be converted into elements.

ELEMENT CLASS
QUAD(4)
RETURN

CONVERT
DIVISIONS : 8, 4
(GEOMETRY/MESH) SURFACES TO ELEMENTS

Select SURFACE.
Close with END LIST.

After defining the element mesh, SWEEP has to be used to remove coinciding
nodes and elements. It is recommended to go to the CHECK menu and check
whether there are elements INSIDE OUT or UPSIDE DOWN. When this is the case
these elements must be fliped.
It is recommended to prevent future drawing of CURVES and SURFACES in the
PLOT-menu.

(MAIN MENU)
BOUNDARY CONDITIONS

Prescribe the MECHANICAL boundary conditions.

Besides the prescribed boundary conditions symmetry conditions have to be prescribed.

Selecting nodes after (NODES) ADD is done with the mouse. Hold the left mouse
button and draw a box around the nodes to be selected. This is easy in this
case, because the nodes are located on a straight line. Do some experiments
with making boxes, by pressing the Control key during box drawing. Close the
selection with END LIST.

(MAIN MENU)
MATERIAL PROPERTIES

Prescribe the material parameters : E = 2.1e11 [Nm2 ] and = 0.3 [-].

The initial yield stress does not have to be prescribed. Mentat uses a default
value, which is very high (v0 = 1020 [Pa]).

12
(MAIN MENU)
GEOMETRIC PROPERTIES
PLANAR
NEW
PLANE STRESS
THICKNESS : 0.01
OK
(ELEMENTS) ADD
(ALL) EXIST

Model difinition is now completed. We are going to specify the analysis.

First we choose the element type. Element 3 is a plane stress element with 4
nodes. (QUAD(4)). As can be seen from the list, there are more elements with
straight edges and four nodes.

(MAIM MENU) (ANALYSIS)

JOBS
ELEMENT TYPES
MECHANICAL
PLANE STRESS
(FULL INTEGRATION) 3
OK
(ALL) EXIST.
RETURN

After choosing the element type, the mechanical (MECHANICAL) analysis is spec-
ified further.

NEW
MECHANICAL

- INITIAL LOADS : all applies

- JOB RESULTS : stress, strain, von-mises
- (ANALYSIS DIMENSION) : PLAIN STRESS
MAIN

SAVE the model as plaatgat1.

13
 The model can now be analysed. This is done by the finite element program
MARC, by submitting the model in (MAIN MENU)(ANALYSIS) JOBS.

(ANALYSIS)
JOBS
RUN
SUBMIT 1
MONITOR

The program MARC is started and the model is analysed. In the RUN
menu we see some information about the analysis. In the status screen the
word Running is seen. When the status indicates Ready the analysis is
finished. When everything has worked well, we see in the RUN menu the exit
number 3004. After completion of the analysis, three fils are written by MARC :

plaatgat1 job1.log
plaatgat1 job1.out
plaatgat1 job1.t16

The results of the analysis can be visualised with Mentat.

The file with extension .out contains the results in alpha-numerical format (ASCII). It is generaly
a rather long file and is mostly only opened when an error has occured during the analysis. The
file with extension .t16 contains the results which can be visualised and post-processed in Mentat.
The file with the extension .log contains information about the analysis.

14
2.2.2 Results

We take a look at the analysis results in Mentat. The .t16 file must be opened.
This can be done directly from within the RUN-menu. It can also be opened
from within the MAIN-menu in the submenu RESULTS. When MARC has been
started by Mentat (via RUN), the Post file can be opened with OPEN DEFAULT.
When Mentat has been closed and we want to open a new Post file, we have
to use the OPEN-button.

(MAIN MENU) (POSTPROCESSING)

RESULTS
OPEN DEFAULT
FILL
DEF & ORG

Because the deformation is very small as it should be for a linear elastic

analysis , we have to enlarge it in SETTINGS and when AUTOMATIC is used the
scaling is slected so that a proper deformation is visible.
Values of the variables selected in JOB RESULTS can be visualised. Selection is
done with SCALAR, TENSOR or VECTOR.

SCALAR
Equivalent Von Mises Stress
OK
CONTOUR BANDS

It is possible to make a plot of a variable along a path in the model, a PATH

PLOT. Here we will plot the y-displacement of the top edge of the plate as a
function of the x-distance along that edge. Try other possibilities.

PATH PLOT
NODE PATH
Enter first node in Path-Plot node path : select (lm) node1
Enter next node in Path-Plot node path (1) : select (lm) node2
etc. etc. close with # (END LIST)
VARIABLES
ADD CURVE
Enter X-axis variable : Arc Length
Enter Y-axis variable : Displacement y
FIT
RETURN
MAIN

Close the .t16 file with CLOSE. It is very important to do this because problems
may occur when loading a (new) model into Mentat. After closing the .t16
file, the model file is restored automatically.

15
2.2.3 Dynamic boundary conditions
Instead of prescribing the displacement of the right edge of the plate, we can also apply a dis-
tributed load.

If necessary, we load the model plaatgat1 in FILES with RESTORE or OPEN. The
prescribed displacement is replaced with a prescribed edge load, which is a
force per unit of area. In BOUNDARY CONDITIONS we make a new apply EDGE
LOAD. To use it we have to select it in INITIAL LOADS. In that case we remove
the prescribed displacement apply3.

FILES
RESTORE or OPEN plaatgat1
MAIN

(MAIN MENU) (PREPROCESSING)

BOUNDARY CONDITIONS
MECHANICAL
NEW
EDGE LOAD
PRESSURE
Enter value for p : -1e8
OK

The given value must be negative for a tensile load.

(EDGES) ADD
Enter add apply element edge list : selecteer EDGES

Remove appl3 in INITIAL LOADS and select apply4.

SAVE the model as plaatgat2, run MARC and look at results.

16
2.3 Axle bearing with radial load
An axle is fixed into a rubber ring in a rigid block as is shown in the figure below. Relevant
dimensions are indicated.

FR

FR

0.05 [m]

0.15 [m]
y
x

0.1 [m]

Material properties of the axle material are known :

Youngs modulus : E = 2.1 1011 [Nm2 ]

Poissons ratio : = 0.3 [-]

The rubber material is assumed to be linearly elastic with the next material parameters given :

Youngs modulus : E = 1.0 107 [Nm2 ]

Poissons ratio : = 0.49 [-]

The axle is loaded by two radial forces, which are equal : FR = 15000 [N]. Bending of the axle is
not taken into account.
It is assumed that a plane strain deformation state exists in the rubber material. Due to
symmetry in the yz-plane w.r.t. the y-axis, only one half of the axle and rubber ring has to be
modelled.
Which half part are you going to model?

17
2.3.1 Modelling and analysis
(MAIN MENU)(PREPROCESSING)
MESH GENERATION

The distance between the grid points is choosen in accordance with the dimen-
sions of the model. In the grid points we can locate model POINTS.

(COORDINATE SYSTEM)
SET
U DOMAIN : -0.1, 0.1
U SPACING :
0.005
V DOMAIN : -0.1, 0.1
V SPACING : 0.005
GRID
FILL
RETURN

We define three ARCS. Because we want to model a solid axle, we define the
third arc of the type CENTER/POINT/POINT with radius zero by clicking three
times on the central grid point (0, 0, 0). This virtual curve can be used to
define a surface.

CURVE TYPE
(ARCS) CENTER/POINT/POINT
RETURN
(CRVS) ADD

Define the half circle of the axle.

Define the half circle of the rubber ring.
Define the central arc with radius zero.
Define the surface between the two half-circles.
Define the surface between the axle-circle and the arc in the center point.

The two defined SURFACES are converted to 4-node elements. The axle sur-
face has some elements near the center point which are triangles. This is no
problem : a quad4 element can have two points coincideing.

ELEMENT CLASS
QUAD(4)
RETURN

CONVERT
DIVISIONS : 8, 8
(GEOMETRY/MESH) SURFACES TO ELEMENTS

Select ring surface

Select the axle surface

RETURN

After generating the element mesh we have to use SWEEP and CHECK.

18
 SWEEP
(SWEEP) ALL
(REMOVE UNUSED) NODES
(REMOVE UNUSED) POINTS
RETURN

CHECK
(CHECK ELEMENTS) UPSIDE DOWN

If necessary FLIP ELEMENTS.

RETURN
MAIN

Boundary conditions (note the symmetry), material properties and geometry

parameters must be prescribed.

(MAIN MENU)(PREPROCESSING)
BOUNDARY CONDITIONS
MECHANICAL

Suppress all displacements of the edge of the rubber ring (apply1).

Nodes on the symmetry axis can only have a displacement in y-direction (apply2).
Prescribe the force FR (apply3).

The question is where this POINT LOAD must be applied, i.e. in which node(s).
Must we prescribe a total force of 15000 N or 30000 N in negative y-direction?

(MAIN MENU)(PREPROCESSING)
MATERIAL PROPERTIES

Prescribe the elastic properties of the axle (material1) and the rubber ring (material2
strangely named material4 by Mentat).

(MAIN MENU)(PREPROCESSING)
GEOMETRIC PROPERTIES
PLANAR
NEW
PLAIN STRAIN
THICKNESS : 0.1
OK
(ELEMENTS) ADD
(ALL) EXIST
MAIN

In (ANALYSIS) JOBS we select element 11, a 4-node linear plane strain element.
Then we select the variables, which we want as output in JOB RESULTS and
specify the load in INITIAL LOADS. Finally we indicate that the deformation
state is plane strain.

(MAIN MENU) (ANALYSIS)

JOBS

19
ELEMENT TYPES
MECHANICAL
PLAIN STRAIN SOLID
(PLANE STRAIN FULL INTEGRATION) 11
OK
(ALL) EXIST
OK
RETURN

NEW
MECHANICAL

- INITIAL LOADS : apply1, apply2, apply3

- JOB RESULTS : stress, strain, von-mises
- (ANALYSIS DIMENSION) : PLAIN STRAIN
MAIN

SAVE the model as asblok1 and run MARC. The results, available in the .t16
files can then be loaded and the results can be visualised.

20
2.3.2 Results

Looking at Von Mises contour plots it is immedeately clear that the stresses in
the rubber are much lower than those in the axle. We can visualize the stress
state in the rubber in more detail by making the axle invissible in the SELECT
menu.

SELECT
ELEMENTS

Select the elements of the axle with a box

Close with END LIST

MAKE INVISIBLE

Push MAKE INVISIBLE again to make everything visible.

21
2.3.3 Quadratic elements

The analysis is done using linear elements with four element nodes (QUAD(4)).
More accurate results, mostly with fewer elements, can be reached, using
quadratic elements with eight element nodes. We can adapt the model as-
blok1. First CLOSE the .t16 file.

FILES
RESTORE or OPEN : asblok1
MAIN

(MAIN MENU) (PREPROCESSING)

MESH GENERATION
CHANGE CLASS
QUAD(8)
ELEMENTS
(ALL) EXIST
RETURN

SWEEP
(SWEEP) ALL
(REMOVE UNUSED) NODES
(REMOVE UNUSED) POINTS
MAIN

(MAIN MENU) (PREPROCESSING)

BOUNDARY CONDITIONS

Some boundary conditions must be adapted, as there are no boundary condi-

tions defined in the new nodes.

Adapt the boundary conditions.

(MAIN MEU) (ANALYSIS)

JOBS
ELEMENT TYPES
MECHANICAL
PLANE STRAIN SOLID
(PLANE STRAIN FULL INTEGRATION) 27
OK
(ALL) EXIST
RETURN
MAIN

SAVE the model and run MARC. Results can be visualised in Mentat.

22
2.4 Orthotropic plate
The figure below shows a square plate, which will be loaded in its plane. It can be assumed that
a plane stress state exists in the plate : zz = xz = yz = 0.

2 y

The plates material is a matrix in which long fibres are embedded, which all have the same
orientation along the direction indicated as 1 in the material 1, 2-coordinate system. Both
matrix and fibres are linearly elastic with Youngs modulus and Poissons ratio Em , Ef , m and
f , respectively. The volume fraction of the fibres is V . The angle between the 1-direction and
the x-axis is = 10 degrees. In the 1, 2-coordinate system the material behaviour for plane stress
is given by the next relation between stress and strain components :

11 E1 21 E1 0 11
1
22 = 12 E2 E2 0 22
1 12 21
12 0 0 (1 12 21 )G12 12

The Youngs moduli, Poissons ratios and shear modulus are defined as :
11 22 22 11 12
E1 = ; E2 = ; 12 = ; 21 = ; G12 =
11 22 11 22 12
The material stiffness matrix is symmetric :

21 E1 = 12 E2

which leaves us with four independent material parameters describing this orthotropic behaviour.

The material parameters can be calculated with the next formulas, which are based on the rule-
of-mixtures :
1 V 1V Ef Em
E1 = V Ef + (1 V )Em ; = + E2 =
E2 Ef Em V Em + (1 V )Ef
1 V 1V Gf Gm
21 = V f + (1 V )m ; = + G12 =
G12 Gf Gm V Gm + (1 V )Gf
The next table lists numerical values for material parameters where the above formulas are used
to calculate E1 , E2 , 12 and G12 .

Em = 70 GPa Ef = 500 GPa E1 = 242 GPa E2 = 106.7 GPa

m = 0.4 f = 0.25 21 = 0.34 G12 = 38.5 GPa
V = 0.4 12 = 0.77

23
In Mentat we also have to give values for E3 , G13 , G23 , 31 and 23 . These values are taken to be
the same as the ones for the matrix material.

E3 = E2 = 106.7 GPa ; G13 = G12 = 38.5 GPa ; G23 = Gm = 25 GPa

31 = 21 = 0.34 ; 23 = m = 0.4

The sides of the plate have a length 1 m, and the thickness of the plate is 1 cm. The plate is
modelled in plane stress with 10 10 4-node elements of type 3.

Input of orthotropic material parameters in the material coordinate system can be done with the
following commands.

(MAIN MENU)
MATERIAL PROPERTIES
NEW
ORTHOTROPIC
E11 E1
E22 E2
E33 E3
NU12 12
NU23 23
NU31 31
G12 G12
G23 G23
G31 G31
ADD
(ALL) EXIST.

The orientation of the material coordinate system w.r.t. the global coordinate system must also
be given. This is done per element in the submenu ORIENTATION. In this submenu we see that
there are several (TYPE)s of orientations. The EDGE-types say that the 1-axis of the local coordinate
system is defined by rotation over an ANGLE relative to a particular side of an element. Side 12 can
be seen in the plot of the element mesh as a half-arrow on the element edges. It is the side from
local node 1 to local node 2. We can also define the rotation of the local 1-axis relative to one of
the global coordinate planes.

ORIENTATION
NEW
EDGE12
ANGLE 10
ADD
(ALL) EXIST.

Boundary conditions are defined for tensile loading in x- and y-direction and for simple shear
loading in x-direction. In the latter case the y-displacement of the upper boundary is suppressed.
Tensile loads and shear load are edge loads of 1 GPa.

For all these loadcases, calculate the stress and strain components in the global and the material
coordinate system.

The calculated stress and strain components can be given in the global coordinate system (default)
and in the preferred system. This preferred system coincides with the material coordinate
system which is defined above in the ORIENTATION submenu.

24
2.5 Exercises
2.5.1 Large thin plate with a cirular hole
A large rectangular plate is loaded with a uniform stress xx = as is shown in the figure below.
In the center of the plate is a hole with radius a, much smaller than the dimensions of the plate.

The stresses around the hole can be determined, using an Airy stress function approach. The
relevant stresses are expressed as components in a cylindrical coordinate system, with coordinates
r, measured from the center of the hole, and in the circumferential direction.

a2 a4 a2
    

rr = 1 2 + 1 + 3 4 4 2 cos(2)
2 r r r
2 4
    
a a
= 1 2 1 + 3 4 cos(2)
2 r r
4 2
 
a a
r = 1 3 4 + 2 2 sin(2)
2 r r

The highest stress values are found in points at the hole edge r = a.

(r = a, = 2) = 3 ; (r = a, = 0) =

The Stress Concentration Factor (SCF) is :

max
Kt = =3 [-]

where it is noted that Kt is independent of the hole diameter.

The plate thickness is 0.01 m, and its dimensions in x- and y-direction are 1 m and 0.5 m,
respectively. Numerical values of other relevant parameters are listed in the table below.

a = 0.05 m = 1 kPa
E = 250 GPa = 0.3

Stress components are plotted for = 0 and for = 2 as a function of the radial distance r.

25
 1000 3000
rr
rr
tt
2500 tt
rt rt
500 2000

1500

[Pa]
[Pa]

0
1000

500
500

1000 500
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
r [m] r [m]

Model the plate in Mentat. Model only a quarter of the plate and apply the proper symmetry
conditions. Choose 4-node quadrilateral plane stress elements, 20 in x-direction and 10 in
y-direction. Use a bias factor to refine the element mesh near the hole edge.
(To get a nice element mesh, it might be wise to generate the mesh in two parts.)
Make a PATH PLOT of the stress components rr and as a function of the radial coordinate,
both for = 0 and for = 2 . Compare the results with the analytical solution.
Increase the number of elements using the same bias factor and quantify the relation between
the number of elements and the error in the numerical solution.
Decrease the hole diameter to a = 0.01 m and repeat the analysis.
Use 8-node quadrilateral elements and answer the former questions again.

26
2.5.2 Plate with a square hole
The figure below shows a plate and its dimensions in [mm]. The plate is clamped at the left edge
(x = 0) and loaded with an edge load of 100 N/mm2 at the right edge. The plate is thin, so we
can assume a plane stress state in the material.
Model the upper part of the plate and use the number of elements as indicated in the figure.

dikte = 1
20
2000 N
8 40 20

40 80

4 8
10 10

8 C D 8 5
20
A B E F

The Youngs modulus and Poissons ratio of the material are E = 210000 MPa and = 0.3.

Make TENSOR PLOTS of the maximum, the minimum and the major principal stresses.
(MAIN > RESULTS > MORE > TENSOR PLOT).

Make a PATH PLOT of the Major Stress and the Von Mises Stress along the line ABCDEF .

Model the plate again and replace the square hole by a circular one, as indicated in the figure.
Use approximately the same element mesh as before.

10
100 N/mm2

27
 Make TENSOR PLOTS of the maximum, the minimum and the major principal stresses.
Make a PATH PLOT of the Major Stress and the Von Mises Stress along the line ABCDEF .

The figure below shows the required PATH PLOTs, where x indicated the distance along the path.

200 200

mj mj
150 150
vm vm

100 100
[MPa]

[MPa]
50 50

0 0

50 50

100 100

150 150
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
x [mm] x [mm]

28
2.5.3 Thick-walled pressurized cylinder
The figure shows a thick-walled cylinder with inner radius a and outer radius b. The inner pressure
is pi and the external pressure is pe .

pe

b
a pi
r

The radial and tangential stresses are :

p i a2 p e b 2 (pi pe )a2 b2 p i a2 p e b 2 (pi pe )a2 b2
rr = ; = +
b 2 a2 (b2 a2 )r2 b 2 a2 (b2 a2 )r2
For an open ended cylinder, the axial stress is zero : zz = 0. For a closed cylinder the axial stress
is :
p i a2 p e b 2
zz =
b 2 a2
The radial and tangential stresses are the same for both cases. The radial displacement for an
open ended cylinder is :
1 p i a2 p e b 2 1 + (pi pe )a2 b2
u= r +
E b 2 a2 E (b2 a2 )r
The next table lists values for the relevant geometrical and material parameters, where E and
are Youngs modulus and Poissons ratio of the cylinder material. The height of the cylinder is
0.5 m.
a = 0.25 m b = 0.50 m
E = 250 GPa = 0.33
pi = 100 MPa pe = 0 MPa
For these values, stresses and radial displacement are plotted as a function of the radius r.
4
8 x 10
x 10 2
2

rr
1.9
tt
1.5
zz
1.8
1
1.7
u [m]
[Pa]

0.5
r

1.6

0
1.5

0.5 1.4

1 1.3
0.25 0.3 0.35 0.4 0.45 0.5 0.25 0.3 0.35 0.4 0.45 0.5
r [m] r [m]

29
 Model a quarter of the cylinder in Mentat using 4-node quadrilateral plane stress elements and
apply the proper boundary conditions for symmetry and edge load. Use 16 elements in tangential
direction and 8 in radial direction.
Calculate the stresses, plot them in a path plot and compare the values with the analytical
solution.
Make all elements twice as small and compare again the stresses and displacement with the
numerical solution.
Start again with the initial element mesh, however, using now 8-point quadrilateral elements.

30
2.5.4 Rotating engineering parts
The next figure shows a disc with a central hole.

r z

When the hickness t is uniform not depending on the radius r , the stress components have to
satisfy the next equilibrium equation is :
rr rr
+ + 2 r = 0
r r
where is the density and the constant angular velocity. Using Hookes law for the linear
elastic material behaviour and the strain-displacement relations, this can be transformed into a
differential equation for the radial displacement u :

d2 u 1 du 1 2
 
u
+ 2+ 2 r = 0
dr2 r dr r E

The general sulution is :

1 2 2 r3
 
a2
u = a1 r +
r E 8

where a1 and a2 are integration constants.

   
c2 3+ 2 2 c2 1 + 3
rr = c1 2 r ; = c1 + 2 2 r2 ; zz = 0
r 8 r 8

where c1 and c2 are again integration constants, depending on a1 and a2 .

Solid disc with constant thickness

For a disc with outer radius b, the displacement at r = 0 must remain finite, while the stress
boundary condition is rr (r = b) = 0. Integration constants can now be determined :
 
3+
c1 = 2 b2 ; c2 = 0
8

The stresses are :

 2
 
3+
2 (b2 r2 ) = (3 + )b2 (1 + 3)r2

rr = ; ; zz = 0
8 8

The disc has an outer radius of 0.5 m and a thickness of 0.05 m. It rotates with an angular velocity
of 6 cycles per second. Youngs modulus is E = 200 GPa and Poissons ratio is = 0.3. The mass
density is = 7500 kg/m3 .

31
The rotation is modelled as a CENTRIFUGAL LOAD in BOUNDARY CONDITONS.

(MAIN) BOUNDARY CONDITIONS

MECHANICAL
CENTRIFUGAL LOAD
ANGULAR VELOCITY (CYCLES/TIME) 6
(AXIS OF ROTATION)
X1 0
Y1 0
Z1 0
X2 0
Y2 0
Z2 1
OK
(ELEMENTS) ADD
(ALL) EXIST.

Stress components for the analytical solution are plotted against the radius in the figure below.
5
x 10
12

rr

tt
10

8
[Pa]

0
0 0.1 0.2 0.3 0.4 0.5
r [m]

Model the disc with plane stress elements and calculate the stresses. Present the radial and the
tangential stress components in a path plot.

Constant thickness disc with a central hole

For a disc with outer radius b and a central hole with radius a, the stress boundary conditions are
rr (r = a) = rr (r = b) = 0. Solving the integration constants leads to :
3+ 3+
c1 = 2 (a2 + b2 ) ; c2 = 2 a2 b2
8 8

32
 Model the disc from the former exercise with a central hole having a radius of 0.25 m and calculate
again the radial and tangential stress components. Use plane stress elements and represent the
stress values in a path plot.

Discs with variable thickness

For a disc with variable thickness t(r) the next equilibrium equation in radial direction can be
derived :
(t(r)rrr )
t(r) + 2 t(r)r2 = 0
r
ta a
The disc has inner and outer radius a and b, respectively, and a thickness distribution t(r) = 2 r.
The general solution for the stresses is :
2c1 d1 2c2 d2 3 + 2 2 2c1 2c2 1 + 3 2 2
rr = r + r r ; = d1 rd1 + d2 rd2 r
ata ata 5+ ata ata 5+
with
q q
d1 = 21 + 5
4 + ; d2 = 21 5
4 +

The integration constants can be determined from the boundary conditions rr (r = a) = rr (r =

b) = 0.

33
3 Axisymmetry
3.1 Background : Axisymmetric modelling
Many devices and components have a geometry which is symmetric w.r.t. an axis and are thus
called axisymmetric. It is obvious that the position of material points of such an object can best
be described in a cylindrical coordinate system. Coordinates are the distance z measured along
the axis of symmetry, the distance r from and perpendicular to that axis and an angle , which
indicates the position in circumferential direction, with respect to an arbitrary starting point
( = 0o : 0 360o(= 2[rad]) (see figure below).

z z

r r
r

When the load is independent of the angle , it is also called axisymmetric. With the additional
assumption that there is no rotation around the z-axis, we only have to consider and model the
cross-section of the object, when we want to analyse its mechanical behaviour (see figure).
Modelling such axisymmetric problems in MARC, the z-axis is oriented in the x-direction and
the r-axis in the y-direction. Only one half of the geometry of the cross-section must be modelled
and it must be located in the half-space y > 0, as is shown in the figure below.

z y = r-as

x = z-as

r r

The stress state in a material point is charaterised by four stress components, which are indicated
on the faces of a stress cube in the figure below. This stress cube is shown two times, once in the
cross-section and once three-dimensional.

34
The deformation is described by four strain components, which are defined in accordance with the
stress components. For linear elastic material behaviour we have :

rr 1 0 rr
tt E 1 0 tt
zz = (1 + )(1 2) 0

0 1 0 zz
rz 0 1 2 rz

rr
rz
rr
y = r-as zr
rz tt zz
zr
P
tt zz

x = z-as

Analysing axisymmetric problems with the finite element method implies that axisymmetric ele-
ments must be used, which are defined in the cross-section. The deformation of such an element is
defined by the displacement in r- and z-direction of the element nodal points. We must be aware
that a nodal point is in fact a nodal ring as is indicated in the figure below.

y = r-as

x = z-as

In the cross-section of the element the axial and radial displacement of a point is interpolated
betwen the displacements of the element nodes. As with the plane elements, linear or quadratic
interpolation can be used. Again we have 4-node and 8-node elements with 4 and 9 integration
points respectively. In MARC these elements are indicated as element type 10 and 28. The
cross-section of a 4-node element is shown in the figure below.

35
 3
4
4 3

1 2 1 2

3
4
4
3
1

3 3
2
1
1 2
1

3 3

In MARC/Mentat the strains in the integration points are defined as follows :

strain 1 = global zz-strain = zz

strain 2 = global rr-strain = rr
strain 3 = global tt-strain = tt
strain 4 = global rz-strain = rz

Stress components are defined in the same way.

36
3.2 Axle bearing with axial load
The load on the axle is assumed to be an axial force : FA = 20000 [N], as is indicated in the figure
below.

FA

y
x

The geometry and the load allow for an axisymmetric modelling and analysis, as indicated in the
figure below.

manchet
y
x as FA

NB. : The x-axis is the axial axis and the positive y-axis is the radial axis. This is the
usual definition in Mentat/MARC.

37
3.2.1 Modelling, analysis and results
(MAIN MENU)(PREPROCESSING)
MESH GENERATION

Define the Cartesian coordinate system with grid point spacing 0.005 [m].

CURVE TYPE
LINE
RETURN
(CRVS) ADD

Define three LINES parallel to the x-axis between 0.05 < x < 0.05 for y = 0, y =
0.025 and y = 0.075.
Define SURFACES between these CURVES.
Convert the surfaces in elements of type QUAD(4) with DIVISIONS [8, 4].
SWEEP and CHECK the mesh.

(MAIN MENU) (PREPROCESSING)

BOUNDARY CONDITIONS
MECHANICAL

Fix the top edge of the rubber part (apply1).

Nodes on the axial axis (y = 0) can only have a displacement in the x-direction (=
axial) (apply2).
Apply the axial force in one point of the axle (apply3).

Material properties of axle and rubber are presribed in the usual way. Geo-
metric parameters are not relevant for this axisymmetric analysis.

In (ANALYSIS) JOBS we select element 10, a 4-node quadrilateral axisymmetric

element. We indicate that the analysis is axisymmetric.

(MAIN MENU) (ANALYSIS)

JOBS

ELEMENT TYPES
MECHANICAL
AXISYMMETRIC SOLID
(FULL INTEGRATION) 10
OK
(ALL) EXIST
RETURN
RETURN

NEW
MECHANICAL

- INITIAL LOADS : apply1, apply2, apply3

- JOB RESULTS : stress, strain, von-mises
- (ANALYSIS DIMENSION) : AXISYMMETRIC
MAIN

38
 SAVE the model as asblok2 and run MARC. The results, available in the .t16
file can be loaded into Mentat and visualised.

Change the model to use quadratic 8-node elements. For axisymmetric analyses in MARC this is
element 28.

39
3.3 Exercises
3.3.1 Circular plate with a circular hole
The next figure shows a plate with a central hole.

a b

The radial displacement and the radial and tangential stresses are given as a function of the radius
r:
c2
u r = c1 r +
r
E h c2 i
rr = 2
(1 + )c1 (1 ) 2
1 r
E h c2 i
tt = (1 + )c1 + (1 ) 2
1 2 r
where E is Youngs modulus and is Poissons ratio. The integration constants c1 and c2 are
determined by the boundary conditions. For ur (r = b) = ub ; rr (r = a) = 0 we have :

(1 )b (1 + )a2 b
c1 = ub ; c2 = ub
(1 )b2 + (1 + )a2 (1 )b2 + (1 + )a2

The next parameter values are given :

a = 0.01 [m] ; b = 0.1 [m] ; d = 0.01 [m]

E = 2 1011 [Pa] ; = 0.3
ub = 0.001 [m]

Make an axisymetric model and calculate the deformation and stress state in the plate.
Compare the results to the exact solution.

The above model was axisymmetric and axisymmetric elements were used in the analysis (see
figure below).

40
 y = radial
x = axial
z = tangential

It is also possible to use a plane stress or plane strain model to calculate the deformation and the
stresses. In this case we model and analyse a quarter of the plate as shown in the figure below.

y
x
x
z

To prescribe the radial displacement on the outer edge, a local coordinate system has to be used,
which can be defined in TRANSFORMS.

(MAIN MENU) (PREPROCESSING)

BOUNDARY CONDITIONS
MECHANICAL
TRANSFORMS
NEW
CYLINDRICAL

Give two points to define the axis of the cylindrical coordinate system, eg. [0 0 0] and
[0 0 1].

(NODES) ADD select nodes at outer edge

MAIN

Make a plane stress model and calculate the deformation and stress state in the plate.
Compare the results to the axisymmetric solution.

41
In the following two figures the stresses from a plane stress analysis are plotted for the planar and
the axisymmetric model.
8 8
x 10 x 10
6 6

rr rr
tt tt
5 5

zz zz

4 4

3 3
[Pa]

[Pa]
2 2

1 1

0 0

1 1
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
r [m] r [m]

3.3.2 Rotating engineering parts

Model the rotating discs from the plane stress exercises with axisymmetric elements and calculate
the stresses.
Present the radial and the tangential stress components in a path plot.

42
4 Three-dimensional models
4.1 Background : Three-dimensional elements
Each component (or structure) in real life has obviously dimensions in three coordinate directions
and is thus a three-dimensional solid object. In previous sections special cases were considered
where dimensions and load were such that a two-dimensional model could be used : plane stress,
plane strain or axisymmetry.
When such a simplification is not allowed, a complete three-dimensional model must be made
and analyzed. In that case the object is subdivided in three-dimensional elements : hexahedrals
(bricks) and/or tetrahedrons (pyramids). Here, we will consider only brick elements, either with
8 nodes in the corners or with 20 nodes in corners and midpoints of the edges (see figure below).
8 8
16 15
7 7
5 5 20
6 6
13 14
17 19
4 12 4
18 11
1 3 1 3
9 10
2 2

In MARC these element have type numbers 7 and 21, respectively. A side of a brick element is
called a face and face loads (forces per unit of area) can be applied on it.
Displacement components of an internal point are interpolated between their values in the nodes
and this interpolation is linear for the 8-node brick and quadratic for the 20-node element. For
integration over the element volume, 8 and 27 integration points are used for the 8- and 20-node
brick, respectively.

Subdividing a three-dimensional solid object in hexagonal elements can be done automatically in

Mentat (or sometimes preferably in Unigraphics). For simple objects and meshes another method
is easily used, which is based on expansion of a two-dimensional mesh in the third dimension.
The Mentat command to do this is EXPAND in the MESH GENERATION menu. We can expand by
translation and/or by rotation as will be shown in the examples.

43
4.2 Three-dimensional model
We will model a rectangular plate, load it with a tensile load and calculate its deformation and
stress state. First we will analyse it under the assumption of plane stress and plane strain. Then
we will expand its geometry to three dimensions and analyse it with boundary conditions, similar
to plane stress and plane strain.

(MAIN MENU)(PREPROCESSING)
MESH GENERATION

Define a coordinate grid with 0.1 u 0.1, 0.1 v 0.1 and u- and v-spacing
0.01. All distances being in meters.

ELEMENT CLASS
QUAD 4
OK
(ELEMS) ADD

Define one element of class QUAD 4 with corner nodes in (0 0 0), (0.1 0 0), (0.1 0.05
0), (0 0.05 0). Subdivide this element in 10 5 elements. SWEEP and CHECK the
mesh.
Boundary conditions are prescribed. The points on the edge y = 0 are not allowed to move in
y-direction. The points on the edge x = 0 are not allowed to move in x-direction. The point
(0, 0, 0) is completely fixed.
The loading is done by a tensile edge load of -1e6 Pa at edge x = 0.1.

Define these boundary conditions.

Material properties for isotropic linear elastic material behaviour are specified : Youngs modulus
E = 210 GPa and Poissons ratio = 0.3.

Because we first want to do a plane anlysis, we have to specify in GEOMETRIC PROPERTIES the
thickness : 0.01 m. First we choose PLANAR PLANE STRESS THICKNESS. For the plane strain analysis
we have to change this to PLANAR PLANE STRAIN THICKNESS.

In the JOBS menu we select the ELEMENT TYPE. First we choose number 3 for plane stress and
afterwards we select number 11 for plane strain.

In JOB RESULTS we choose stress and total strain as output parameters.

FILES SAVE AS plate2D and run the model in MARC. Look up the relevant stresses and strains
in Mentat and repeat this for a plane strain analysis.

Results :
11 11 22
plane stress 1.000e6 4.762e-6 -1.429e-6
plane strain 1.000e6 4.333e-6 -1.857e-6

Now the model plate2D is opened again and adapted to three dimensions. We will save it later
as plate3D.

We go to (MAIN MENU)(PREPROCESSING) and MESH GENERATION and then in the submenu EXPAND,
where we expand the element mesh in the third (z-) direction. The number of element layers is

44
chosen to be 5, with each layer being 0.002 m thick. The total thickness is thus 0.01 m.

(MAIN MENU)(PREPROCESSING)
MESH GENERATION
EXPAND
TRANSLATIONS : 0 0 0.002
REPETITIONS : 5
ELEMENTS
(ALL) EXISTING
RETURN

Now SWEEP the mesh and CHECK for INSIDE OUT elements.

When we rotate the model we will see that we have made a three-dimensional element mesh. In
the BOUNDARY CONDITIONS menu we can visualize all boundary conditions and it becomes clear
that prescribed nodal point displacements have been copied with the nodes. It remains to be seen
if this is what we want. The edge load, however, has disappeared and must be defined again but
now as a face load on the element faces on the right side of the model.

MATERIAL PROPERTIES are also copied. GEOMETRIC PROPERTIES must be removed as they are not rel-
evant for our three-dimensional model.

The only thing which is left to do is selecting the correct element type in JOBS ELEMENT TYPES
3-D SOLID.We choose element 7, which is a 8-node brick element.

Save the model as plate3D and run it in MARC.

When we look at the results, we see that the deformation and stress state differs from those in
the two-dimensional analyses. Obviously this is due to the boundary conditions.

Go back to the plate3D model and make boundary conditions, which are analoguous to plane
stress and plane strain respectively. Compare the stresses and strains with the values from the
two-dimensional analyses.

The same expansion which is described above, can also be applied to make a three-dimensional
model out of an axisymmetric one. In that case instead of TRANSLATIONS we have to use ROTATIONS
around the x-axis, which is the axis of axisymmetry in Mentat/MARC.

Take an axisymmetric model of a disc from he exercises and make a three-dimensional

disc out of it.

45
4.3 Exercise
4.3.1 Inhomogeneous plate with a central circular hole
The figure below shows a square plate (dimensions 0.2 0.2 m) with a central circular hole (radius
= 0.03 m). The plate is made from two different materials A and B. Material A is located between
the hole and a circle with radius 0.07 m. In the undeformed state the plate thickness is 0.04 m
everywhere.

0.2
y
0.07

P
A

Q x

0.03

The plate is axially loaded with an edge load = 50 MPa. A plane stress state is assumed to
result in the plate after loading.

Model a quarter of the plate [ 0 x 0.1 ; 0 y 0.1 ] and apply the correct
boundary conditions.
Make an element mesh with 10 quad4 elements in tangential direction and 20 elements
in radial direction.

The following table lists relevant material parameters.

Youngs modulus material A EA 400 GPa

Poissons ratio material A A 0.25 -
Youngs modulus material B EA 200 GPa
Poissons ratio material B A 0.3 -

What are the x- and y-displacements in points P and Q ?

What is the maximum value of the stress in the material and in which point M (x, y) is this value
found ?

46
Besides the applied boundary load a new edge load in y-direction is applied on the edges [0.1 <
x < 0.1 ; y = 0.1].

For which value of the new edge load is the stress in point M zero.

Convert the two-dimensional model into a three-dimensional model, using the EXPAND command.
Take 10 elements over the thickness of the plate.

Calculate strains and stresses in the plate material.

Make a PATH PLOT of the strain zz over the thickness in the point with coordinates x = 0.03
and y = 0.
Make a PATH PLOT of the stress zz over the thickness in the point with coordinates x = 0.03
and y = 0.
What can you conclude concerning the assumption of plane stress ?

47