8

Finite Element
Modeling:
Mesh and BCs

8–1
Chapter 8: FINITE ELEMENT MODELING: MESH AND BCS 8–2

§8.1 INTRODUCTION
This Chapter continues the exposition of FEM modeling principles. It provides guidelines on
the layout of finite element meshes, the conversion of distributed loads to node forces, and on
handling the simplest forms of support boundary conditions. The next two Chapters deal with more
complicated forms of boundary conditions called multifreedom constraints.
The presentation is “recipe oriented”, illustrated by specific examples. All examples are from
structural mechanics and most of them are two-dimensional. No attempt is given at a rigorous
justificatiuon of rules and recommendations, as that would require mathematical tools beyond the
scope of this treatment.

§8.2 GUIDELINES ON ELEMENT LAYOUT

The following guidelines are stated for structural applications. As noted above, they will be often
illustrated for two-dimensional meshes of continuum elements for ease of visualization.

§8.2.1 General Recommendations

The general rules that should guide you in the use of commercial or public FEM packages, are:

1. Use the simplest type of finite element that will do the job.
2. Never, never, never mess around with complicated or special elements, unless you
are absolutely sure of what you are doing.
3. Use the coarsest mesh you think will capture the dominant physical behavior of the
physical system, particularly in design applications.

Three word summary: keep it simple. Initial FE models may have to be substantially revised to
accommodate design changes, and there is little point in using detailed models that will not survive
design iterations. The time to use refined models is when the design has stabilized and you have a
fairly good picture of the underlying physics.

§8.2.2 Coarse and Fine FE Meshes

Use a relatively fine (coarse) discretization in regions where you expect a high (low) gradient of
strains and/or stresses. Regions to watch out for high gradients are:
1. Near entrant corners, or sharply curved edges.
2. In the vicinity of concentrated (point) loads, concentrated reactions, cracks and cutouts.
3. In the interior of structures with abrupt changes in thickness, material properties or cross
sectional areas.
The examples in Figure 8.1 illustrate some of these “danger regions.” Away from such regions one
can use a fairly coarse discretization within constraints imposed by the need of representing the
structural geometry, loading and support conditions reasonably well.

8–2
8–3 §8.2 GUIDELINES ON ELEMENT LAYOUT

entrant corners
cutouts cracks

abrupt thickness
load transfer changes
Vicinity of concentrated (bonded joints,
(point) loads welds, etc.)

Figure 8.1. Some situations where a locally refined

finite element discretization is recommended.

§8.2.3 Interfaces
A physical interface, resulting from example from a change in material, should also be an interele-
ment boundary. That is, elements must not cross interfaces.

§8.2.4 Element Aspect Ratios

When discretizing two and three dimensional problems, try to avoid finite elements of bad aspect
ratio: elongated or “skinny” elements, as the ones illustrated in Figure 8.2. As a rough guideline,
elements with aspect ratios exceeding 3 should be viewed with caution and those exceeding 10 with
alarm. Such elements will not necessarily produce bad results — that depends on the loading and
boundary conditions of the problem — but do introduce the potential for trouble.

REMARK 8.1
In many “thin” structures modeled as continuous bodies the appearance of “skinny” elements is inevitable on
account of computational economy reasons. An example is provided by the three-dimensional modeling of
structural composites in aerospace and mechanical engineering problems.

§8.2.5 Triangles or Quadrilaterals?

In two-dimensional FE modeling, if you have a choice between triangles and quadrilaterals with
similar nodal arrangement, prefer quadrilaterals. Triangles are quite convenient for mesh generation,
mesh transitions, rounding up corners, and the like. But sometimes triangles can be avoided
altogether with some thought. One of the homework exercises is oriented along these lines.

8–3
Chapter 8: FINITE ELEMENT MODELING: MESH AND BCS 8–4

Good Bad

Figure 8.2. Good and bad element aspect ratios.

§8.2.6 Bricks, Wedges or Tetrahedra?

In three dimensional FE modeling, prefer strongly bricks over wedges, and wedges over tetrahedra.
The latter should be used only if there is no viable alternative. The main problem with tetrahedra is
that they can produce totally wrong stress results even if the displacement solution looks reasonable.

§8.3 TREATMENT OF DISTRIBUTED LOADS

In practical structural problems, distributed loads are more common than concentrated (point) loads,
because one of the objectives of a good design is to avoid or alleviate stress concentrations produced
by concentrated forces. Distributed loads may be of surface or volume type.
Distributed surface loads (called surface tractions in elasticity theory) are associated with actions
such as wind or water pressure, live loads on bridges, and the like. Volume loads (called body forces
in elasticity) are associated with own weight (gravity), inertial, thermal, prestress or electromagnetic
effects.
Whatever their nature or source, distributed loads must be converted to consistent nodal forces.
These forces eventually end up in the right-hand side of the master stiffness equations.
The meaning of “consistent” can be made precise only through variational arguments. Consequently,
a thorough treatment of this topic will be deferred to Part II. However, a simpler approach called
tributary load lumping, or simply load lumping, is often used by structural engineers in lieu of
the more mathematically impeccable but complicated variational approach. The procedure is
graphically illustrated in Figure 8.3 for a distributed surface loading on a two-dimensional FE
mesh, and it is readily extended to three-dimensional meshes as well as volume loads.
This simple lumping procedure should be restricted, however, to elements that have only corner
nodes and to near-uniform load distributions. It does not generally yield correct results if the
elements have additional nodes on their sides or faces. For these “refined” elements the variational
approach is preferable.
Similarly, if the loads vary rapidly (within the element length scale) the lumping method should be

8–4
8–5 §8.5 SUPPORT CONDITIONS

Nodal force P has

magnitude of shaded area
Distributed load
under load curve. This area
intensity (load acts
goes halfway over adjacent
downward on boundary)
element sides

P
Boundary

Finite element
mesh

Figure 8.3. Nodal force “lumping” of surface load.

used with caution for coarse meshes.

§8.4 BOUNDARY CONDITIONS

Boundary conditions (BCs) are discussed from a more general standpoint in later Chapters. The
key distinction made in the previous Chapter between essential and natural The topic is treated
there using a variational setting.
In the present Chapter we discuss next the simplest essential boundary conditions in structural
mechanics from a physical standpoint. This makes them relevant to the types of problem with
which a structural engineer is familiar. Because of the informal setting, the following discussion
relies heavily on examples.
In structural problems, the previous Chapter recipe that distinguishes between essential and natural
B.C.s is: if it directly involves the nodal freedoms, such as displacements or rotations, it is essen-
tial. Otherwise it is natural. Conditions involving applied loads are natural. Essential BCs take
precedence over natural BCs.
The simplest essential boundary conditions are ground or support constraints, and symmetry con-
ditions. These appear in the overwhelming majority of practical problems. More exotic types, such
as multifreedom constraints, require more sophisticated handling techniques and are covered in the
next two Chapters.

§8.5 SUPPORT CONDITIONS

Supports are used to restrain structures against rigid body motions. Consequently, the resulting
boundary conditions are often called motion constraints, or simple constraints. In what follows we
analyze two- and three-dimensional motions separately.

8–5
Chapter 8: FINITE ELEMENT MODELING: MESH AND BCS 8–6


,
,
,
,






,

,
B


,



,
,
,

A

Figure 8.4. Minimal restraints against two-dimensional rigid body motions

§8.5.1 Supporting Two Dimensional Bodies

Figure 8.4 shows an arbitrary two-dimensional body that displaces in the plane of the paper. If
it is not restrained, an arbitrary load will cause infinite displacements. Regardless of the loading
conditions, the structure must be restrained against two translations and one rotation. Consequently
the minimum number of constraints that has to be imposed in two dimensions is three.
In Figure 8.4, support A provides translational restraint, whereas support B, together with A,
provides rotational restraint. In the finite element terminology, we say that we delete all translational
degrees of freedom (meaning all displacements) at point A, and that we delete the translational
degree of freedom directed along the normal to the AB direction at point B. Thus the body is free
to distort in any manner without the supports imposing any displacement constraints.

y
,,,

,
,,,,, A

,
x

Figure 8.5. A special case of Figure 8.4.

A and B are reaction-to-ground points. That is, the applied loads are automatically balanced by
reactive forces at points A and B, for the structure is attached to its “ground” environment at

8–6
8–7 §8.6 TAKING ADVANTAGE OF SYMMETRY CONDITIONS

those points. Additional freedoms may be deleted to model greater restraint by the environment.
However, Figure 8.4 does illustrate the minimal number of constraints.
Figure 8.5 shows a slightly simplified version of Figure 8.4. Here the line AB is parallel to the
global y axis. We simply delete the x and y translations at point A, and the x translation at point B.

,,
y

,,
,,
B

, ,
,,,, ,,,
, ,
,,
D

,,
,,
,,
,,,,,
,,,
A C x

z
Figure 8.6. Suppressing RBM freedoms in a three-dimensional body.

§8.5.2 Supporting Three Dimensional Bodies

Figure 8.6 illustrates the extension of the freedom deletion concept to three dimensions. The minimal
number of freedoms that have to be deleted is now six and many combinations are possible. In the
example, we have chosen to delete all three degrees of freedom at point A to prevent rigid body
translations. The x displacement component at point B is deleted to prevent rotation about z, the
z component is deleted at point C to prevent rotation about y, and the y component is deleted at
point D to prevent rotation about x.

§8.6 TAKING ADVANTAGE OF SYMMETRY CONDITIONS

The structural engineer doing finite element analysis should always be on the lookout for conditions
of symmetry or antisymmetry. Judicious use of these conditions allows only a portion of the structure
to be analyzed, with a consequent saving in data preparation and computer processing time.

§8.6.1 Visualization
The first thing that the analyst must learn is how to recognize these conditions. The technique
recommended here relies on kinetic visualization of the displacement field. It is illustrated for the
two-dimensional case.
A symmetry line in two-dimensional motion can be recognized by remembering the “mirror” dis-
placement pattern shown in Figure 8.7(a). Similarly, an antisymmetry line is characterized by the
pattern illustrated in Figure 8.7(b).

8–7
Chapter 8: FINITE ELEMENT MODELING: MESH AND BCS 8–8

c Symmetry Antisymmetry
(a) line (b) line

A' A A" A' A A"

displacement
vectors

Figure 8.7. Visualizing symmetry and antisymmetry lines

through two-dimensional motion.

Similar recognition patterns can be drawn in three dimensions to help visualization of planes of
symmetry or antisymmetry. More complex regular patterns associated with sector symmetry (also
called harmonic symmetry) and rotational symmetry can be treated in a similar manner, but will
not be discussed here.

§8.6.2 Effect of Loading Patterns

Although the structure may look symmetric in shape, it must be kept in mind that the simplification
can be used only if the loading conditions are also symmetric or antisymmetric.
Consider the plate structure shown in Figure 8.8(a). This structure is symmetrically loaded on the
x-y plane. On applying the recognition patterns shown previously one concludes that the structure
is doubly symmetric in both geometry and loading. It is evident that no displacements in the x-
direction are possible for any point on the y-axis, and that no y displacements are possible for points
on the x axis. A finite element model of this structure may look like that shown in Figure 8.8(b).
On the other hand if the loading is antisymmetric, as shown in Figure 8.9(a), then the x axis
becomes an antisymmetry line because none of the y = 0 points can move along the x direction.
The boundary conditions to be imposed on the finite element model are also different, as shown in
Figure 8.9(b).

REMARK 8.2
For the antisymmetric loading case, one node point has to be constrained against vertical motion. The choice
is arbitrary and amounts only to an adjustment on the overall (rigid-body) vertical motion. In Figure 8.9(b) the
center point C has been chosen to be that vertically-constrained node. But any other point could be selected
as well; for example A or D.

8–8
8–9 §8.6 TAKING ADVANTAGE OF SYMMETRY CONDITIONS

y
(a)
L
c
A B

C D x
L
c

(b) ,, A B

,,, C D
,
,,
,
Figure 8.8. A doubly symmetric structure under symmetric loading.

y
(a)
A
L
c
B

x
C
D L
A

(b) ,, A B

,,,,
C
D

Vertical (y) motion

of C may be constrained

Figure 8.9. A doubly symmetric structure under antisymmetric loading.

8–9
Chapter 8: FINITE ELEMENT MODELING: MESH AND BCS 8–10

Homework Exercises for Chapters 7 and 8

EXERCISE 8.1
The plate structure shown in Figure E8.1 is loaded and deforms in the plane of the figure. The applied load at
D and the supports at I and N extend over a fairly narrow area. Give a list of what you think are the likely
“trouble spots” that would require a locally finer finite element mesh to capture high stress gradients.
Identify those spots by its letter and a reason. For example, D: vicinity of point load.

P
C E
A D
B F G

,, ,,
N M J I
O H
L K

Figure E8.1. Plate structure for Exercise 8.1.

EXERCISE 8.2
Part of a two-dimensional FE mesh has been set up as indicated in Figure E8.2. Region ABC D is still
unmeshed. Draw a transition mesh within that region that correctly merges with the regular grids shown, uses
4-node quadrilateral elements (quadrilaterals with corner nodes only), and avoids triangles. Note: There are
several (equally acceptable) solutions.

A B

D C

Figure E8.2. Plate structure for Exercise 8.2.

EXERCISE 8.3
Compute the “lumped” nodal forces p1 , p2 , p3 and p4 equivalent to the linearly-varying distributed surface load
q for the finite element layout defined in Figure E8.3. For example, p1 = 3q/8. Check that p1 + p2 + p3 + p4 =
6q (why?). Note that q is given as a force per unit of vertical length.

8–10
8–11 Exercises

6
2
4
3
2
4 q

Figure E8.3. Mesh layout for Exercise 8.3.

P P
(a) (b)
P P

P P
,, ,,

P (d) (uniform)
(c)
45o

fixed
q center q
hole
45o

(e) P (f) P P

infinite half plane

Figure E8.4. Problems for Exercise 8.4.

8–11
Chapter 8: FINITE ELEMENT MODELING: MESH AND BCS 8–12

EXERCISE 8.4
Identify the symmetry and antisymmetry lines in the two-dimensional problems illustrated in Figure E8.4.
They are: (a) a circular disk under two diametrically opposite point forces (the famous “Brazilian test” for
concrete); (b) the same disk under two diametrically opposite force pairs; (c) a clamped semiannulus under a
force pair oriented as shown; (d) a stretched rectangular plate with a central circular hole. Finally (e) and (f)
are half-planes under concentrated loads.1
Having identified those symmetry/antisymmetry lines, state whether it is possible to cut the complete structure
to one half or one quarter before laying out a finite element mesh. Then draw a coarse FE mesh indicating, with
rollers or fixed supports, which kind of displacement BCs you would specify on the symmetry or antisymmetry
lines. Note: Do all sketches on your paper, not on the printed figures.

EXERCISE 8.5
You (a finite element guru) pass away and come back to the next life as an intelligent but hungry bird. Looking
around, you notice a succulent big worm taking a peek at the weather. You grab one end and pull for dinner;
see Figure E8.5.
B

, , , , , , ,
, ,
,
G

,,,,,,,, E
D
friction force on
worm surface

Figure E8.5. The hungry bird.

After a long struggle, however, the worm wins. While hungrily looking for a smaller one your thoughts wonder
to FEM and how the bird extraction process might be modeled so you can pull it out more efficiently. Then
you wake up to face this homework question. Try your hand at the following “worm modeling” points.
(a) The worm is simply modeled as a string of one-dimensional (bar) elements. The “worm axial force” is
of course constant from the beak B to ground level G, then decreases rapidly because of soil friction
(which varies roughly as plotted in the figure above) and drops to nearly zero over D E. Sketch how a
good “worm-element mesh” should look like to capture the axial force well.
(b) On the above model, how would you represent boundary conditions, applied forces and friction forces?
(c) Next you want a more refined anaysis of the worm that distinguishes skin and insides. What type of
finite element model would be appropriate?

1 Note that (e) is the famous Flamant’s problem, which is important in the 2D design of foundations of civil structures.
The analytical solution of (e) and (f) may be found, for instance, in Timoshenko-Goodier’s Theory of Elasticity, 2nd
Edition, page 85ff.

8–12
8–13 Exercises

(d) (Advanced) Finally, point out what need to be added to the model of (c) to include the soil as an elastic
medium.
Briefly explain your decisions. Avoid writing equations.

8–13