Lecture

32: FE Mesh Genera3on

APL705 Finite Element Method

 FE Stages


•  The en3re finite element process can be divide into three
main groups of ac3vi3es
•  1. Pre-processing
•  2. Formula3on and solu3on
•  3. Post-processing
•  The first part involves the geometry crea3on and
discre3za3on of the problem domain
•  The bulk of the second stage is concerned with element
formula3on, assembly of s3ffness matrices and the most
important numerical solu3on of the system of equa3ons.
•  In the last stage, we present and analyse the results obtained
from stage two.
 Preprocessing


•  Geometry crea3on for the problem domain
•  Crea3on of the nodal and elemental data, coordinates,
connec3vity, boundary condi3ons, loading data and material
proper3es
•  This ac3vity is known by a common name ‘Mesh Genera3on’.
•  As the number of elements and nodes grow in real world
problems, it becomes unmanageable to handle such data in
any stage of FEM.
•  Therefore in modern FE soTware programs, we find very
evolved Pre-processors / mesh generators which facilitate
modeling and solving complex problems.

 Mesh Genera3on


•  The basic purpose of mesh genera3on process is to create element
connec3vity data and nodal coordinates by taking some key points
as inputs. For convenience of demonstra3on we can consider a two
dimensional domain.
•  Consider the 2D problem domain shown below – a quarter of a
circle. This we call the region.
•  The corresponding
block is shown as
rectangular area
which helps us to
generate node
numbers
•  Note that the regi-
on 4 is a void and
corresponds to a line
in problem domain
 Genera3on of Node Numbers


•  The strategy of node numbering is illustrated using an
example considered previously. The merger of edges 18-20
and 18-28 and arriving at the final node numbering as shown
below is key to a consistent node numbering.
 Genera3on of Coordinates and Connec3vity


•  For the purpose of genera3ng coordinates and connec3vity, we can use

the isoparametric master element developed earlier with 8 nodes and 8
shape func3ons.

•  Other ac3vi3es associated with it are

–  Mesh Plo]ng
–  Data handling and edi3ng
 Post-Processing


•  This is an important task aTer obtaining the results by solving the system

of equa3ons. In most present day soTware programs both pre and post
processors are part of the GUI.
•  Plo]ng of deformed shape / mode shape – Gives an idea of
how the structure behaves under the applied loads
•  Contour plo]ng – different types of results can be graphically
visualised using either symbol plots or colour plots.
•  Displaying nodal and element values of variables
•  Calcula3on and display of secondary variables and their
analysis
 Mesh Genera3on and Tests


•  While genera3ng the mesh, a`en3on is necessary for the
correctness and accuracy of the results to be achieved.
•  It is necessary to consider the convergence aspect of FE
approxima3on right in the beginning
•  Also, rate of convergence becomes important in large and
complex problems
•  A check on the robustness of solu3on procedure and
algorithms
•  The correctness of programming and implementa3on
 Convergence Criteria



•  We know that FE method is an approxima3on of the exact
con3nuum solu3on.
•  The convergence test decides how closely the two solu3ons
agree with each other.
•  The approximate solu3on u’ should tend to the exact solu3on
u when the size of the element h approaches zero, i.e
| u − u' |= O(h q ) ≤ Ch q
•  Where h is sufficiently small, q>0 and C is posi3ve constant.
•  This must also be true for all the deriva3ves of u
•  The value o q indicates the order of convergence
 Convergence Criteria


•  The approxima3on should be such that, it sa3sfies both
consistency and stability condi3ons
| u − u' |= O(h q ) ≤ Ch q
•  The consistency requirements: As the size of the element h
tends to zero the approximate equa3on (model) will
represent the exact differen3al equa3on and boundary
condi3ons.
•  The stability Condi3on: It means that the solu3on of the
discre3zed system be unique and avoid spurious inclusions
that may pollute the solu3on itself.
a = K −1 f
•  Which simply mean that the s3ffness matrix be non-singular
 Patch Test


•  It is a procedure to test the consistency requirements in FEM
•  The basic principle of this test is based on the idea that if we
consider sufficiently small domain of size, 2h, we can expand
the unknown func3on u and its essen3al deriva3ves
appearing in the weak formula3on into a Taylor series
•  Around point i we require that u = u + ⎛⎜ ∂u ⎞⎟ x + ⎛⎜ ∂u ⎞⎟ y +!+ O(h p )
with p≥2
i
⎝ ∂x ⎠i ⎝ ∂y ⎠i
•  The FE approxima3on should ∂u = ⎛⎜ ∂u ⎞⎟ +!+ O(h p−1 )
reproduce exactly the problem ∂x ⎝ ∂x ⎠i
posed for any linear forms of ∂u ⎛ ∂u ⎞
= ⎜ ⎟ +!+ O(h p−1 )
u as h tends to zero ∂y ⎝ ∂y ⎠i
•  The patch test tests only the sa3sfac3on of the differen3al
equa3ons but not the boundary condi3ons.
 Element Quality Checks


•  As we are assuming the solu3on to be approximate, we need
to consider quality of elements in the first place for improving
the accuracy of the solu3on obtained.
•  Some examples are shown
here graphically
•  Common quality checks test
for
•  Skewness, Aspect ra3o,
Warpage and Jacobian ra3o

 Element Quality Parameters


•  Skewness: It is defined as the angle (90-α) between the vectors for
triangle and a quad element
are shown.
•  Aspect ra3o : This is the ra3o of
maximum length of side (dimension) to
to minimum side-length of an element

•  Warpage: warpage of 2D elements is found by dividing a quad
element into 2 triangular elements and finding the angle between
the plane of the two triangles. It is done again with dividing the
other corners. The maximum of the two angles is the warpage.
•  Jacobian: The Jacobian ra3on is the measure of the devia3on of the
element form the ideal shape element (master element). This ra3o
ranges from -1 to 1 where 1 corresponds to an ideal element.
Acceptable value is >0.6