SlideShare a Scribd company logo
Mesh Quality
               Julien Dompierre
          julien@cerca.umontreal.ca


                                     ´
Centre de Recherche en Calcul Applique (CERCA)
        ´                           ´
        Ecole Polytechnique de Montreal




                                                 Mesh Quality – p. 1/331
Authors
•   Research professionals
    • Julien Dompierre
    • Paul Labbé
    • Marie-Gabrielle Vallet
•   Professors
    • François Guibault
    • Jean-Yves Trépanier
    • Ricardo Camarero




                               Mesh Quality – p. 2/331
References – 1

            J. D OMPIERRE , P. L ABBÉ ,
            M.-G. VALLET, F. G UIBAULT
            AND R. C AMARERO , Critères
            de qualité pour les maillages
            simpliciaux. in Maillage et
            adaptation, Hermès, October
            2001, Paris, pages 311–348.




                                  Mesh Quality – p. 3/331
References – 2


A. L IU and B. J OE, Relationship between
Tetrahedron Shape Measures, Bit, Vol. 34,
pages 268–287, (1994).




                                     Mesh Quality – p. 4/331
References – 3


P. L ABBÉ, J. D OMPIERRE, M.-G. VALLET, F.
G UIBAULT and J.-Y. T RÉPANIER, A Universal
Measure of the Conformity of a Mesh with
Respect to an Anisotropic Metric Field,
Submitted to Int. J. for Numer. Meth. in Engng,
(2003).




                                        Mesh Quality – p. 5/331
References – 4


P. L ABBÉ, J. D OMPIERRE, M.-G. VALLET, F.
G UIBAULT and J.-Y. T RÉPANIER, A Measure of
the Conformity of a Mesh to an Anisotropic
Metric, Tenth International Meshing Roundtable,
Newport Beach, CA, pages 319–326, (2001).




                                       Mesh Quality – p. 6/331
References – 5

            P.-L. G EORGE AND H. B O -
            ROUCHAKI , Triangulation de
            Delaunay et maillage, appli-
            cations aux éléments finis.
            Hermès, 1997, Paris.
            This book is available in En-
            glish.




                                  Mesh Quality – p. 7/331
References – 6


            P. J. F REY AND P.-L.
            G EORGE, Maillages. Ap-
            plications aux éléments finis.
            Hermès, 1999, Paris.
            This book is available in
            English.




                                  Mesh Quality – p. 8/331
Table of Contents

1. Introduction        8. Non-Simplicial
2. Simplex Definition     Elements
3. Degeneracies of     9. Shape Quality
  Simplices              Visualization
4. Shape Quality of    10. Shape Quality
  Simplices              Equivalence
5. Formulae for Sim-   11. Mesh Quality and
  plices                 Optimization
6. Voronoi, Delaunay   12. Size Quality of
  and Riemann            Simplices
7. Shape Quality and   13. Universal Quality
  Delaunay             14. Conclusions
                                   Mesh Quality – p. 9/331
Introduction and Justifications

We work on mesh generation, mesh adaptation
and mesh optimization.




How can we choose the configuration that
produces the best triangles ? A triangle shape
quality measure is needed.
                                       Mesh Quality – p. 10/331
Face Flipping




How can we choose the configuration that
produces the best tetrahedra ? A tetrahedron
shape quality measure is needed.



                                       Mesh Quality – p. 11/331
Edge Swapping
               S4   S3                 S4   S3

          S5                      S5
   A                          A
                          B                                      B

                     S2                          S2


               S1                      S1

How can we choose the configuration that
produces the best tetrahedra ? A tetrahedron
shape quality measure is needed.


                                                      Mesh Quality – p. 12/331
Mesh Optimization

 •   Let O1 and O2 , two three-dimensional
     unstructured tetrahedral mesh Optimizers.




                                        Mesh Quality – p. 13/331
Mesh Optimization

 •   Let O1 and O2 , two three-dimensional
     unstructured tetrahedral mesh Optimizers.
 •   What is the norm O of a mesh optimizer ?




                                        Mesh Quality – p. 13/331
Mesh Optimization

 •   Let O1 and O2 , two three-dimensional
     unstructured tetrahedral mesh Optimizers.
 •   What is the norm O of a mesh optimizer ?
 •   How can it be asserted that O1 > O2 ?




                                        Mesh Quality – p. 13/331
It’s Obvious !

 •   Let B be a benchmark.




                             Mesh Quality – p. 14/331
It’s Obvious !

 •   Let B be a benchmark.
 •   Let M1 = O1 (B) be the optimized mesh
     obtained with the mesh optimizer O1 .




                                       Mesh Quality – p. 14/331
It’s Obvious !

 •   Let B be a benchmark.
 •   Let M1 = O1 (B) be the optimized mesh
     obtained with the mesh optimizer O1 .
 •   Let M2 = O2 (B) be the optimized mesh
     obtained with the mesh optimizer O2 .




                                       Mesh Quality – p. 14/331
It’s Obvious !

 •   Let B be a benchmark.
 •   Let M1 = O1 (B) be the optimized mesh
     obtained with the mesh optimizer O1 .
 •   Let M2 = O2 (B) be the optimized mesh
     obtained with the mesh optimizer O2 .
 •   Common sense says : “The proof is in the
     pudding”.




                                       Mesh Quality – p. 14/331
It’s Obvious !

 •   Let B be a benchmark.
 •   Let M1 = O1 (B) be the optimized mesh
     obtained with the mesh optimizer O1 .
 •   Let M2 = O2 (B) be the optimized mesh
     obtained with the mesh optimizer O2 .
 •   Common sense says : “The proof is in the
     pudding”.
 •   If M1 > M2 then O1 > O2 .


                                       Mesh Quality – p. 14/331
Benchmarks for Mesh Optimization

J. D OMPIERRE, P. L ABBÉ, F. G UIBAULT and
R. C AMARERO.
Proposal of Benchmarks for 3D Unstructured
Tetrahedral Mesh Optimization.
7th International Meshing Roundtable, Dearborn,
MI, October 1998, pages 459–478.




                                      Mesh Quality – p. 15/331
The Trick...

 •   Because the norm O of a mesh optimizer is
     unknown, the comparison of two optimizers is
     replaced by the comparison of two meshes.




                                       Mesh Quality – p. 16/331
The Trick...

 •   Because the norm O of a mesh optimizer is
     unknown, the comparison of two optimizers is
     replaced by the comparison of two meshes.
 •   What is the norm M of a mesh ?




                                       Mesh Quality – p. 16/331
The Trick...

 •   Because the norm O of a mesh optimizer is
     unknown, the comparison of two optimizers is
     replaced by the comparison of two meshes.
 •   What is the norm M of a mesh ?
 •   How can we assert that M1 > M2 ?




                                       Mesh Quality – p. 16/331
The Trick...

 •   Because the norm O of a mesh optimizer is
     unknown, the comparison of two optimizers is
     replaced by the comparison of two meshes.
 •   What is the norm M of a mesh ?
 •   How can we assert that M1 > M2 ?
 •   This is what you will know soon, or you
     money back !




                                         Mesh Quality – p. 16/331
What to Retain

 •   This lecture is about the quality of the
     elements of a mesh and the quality of a whole
     mesh.




                                        Mesh Quality – p. 17/331
What to Retain

 •   This lecture is about the quality of the
     elements of a mesh and the quality of a whole
     mesh.
 •   The concept of element quality is necessary
     for the algorithms of egde and face swapping.




                                        Mesh Quality – p. 17/331
What to Retain

 •   This lecture is about the quality of the
     elements of a mesh and the quality of a whole
     mesh.
 •   The concept of element quality is necessary
     for the algorithms of egde and face swapping.
 •   The concept of mesh quality is necessary to
     do research on mesh optimization.




                                        Mesh Quality – p. 17/331
Table of Contents


1. Introduction             8. Non-Simplicial
2. Simplex Definition          Elements
3. Degeneracies of          9. Shape Quality
  Simplices                   Visualization
4. Shape Quality of         10. Shape Quality
  Simplices                   Equivalence
5. Formulae for Simplices   11. Mesh Quality and
6. Voronoi, Delaunay and      Optimization
  Riemann                   12. Size Quality of
7. Shape Quality and          Simplices
  Delaunay                  13. Universal Quality
                            14. Conclusions



                                          Mesh Quality – p. 18/331
Definition of a Simplex



Meshes in two and three dimensions are made of
polygons or polyhedra named elements.




                                        Mesh Quality – p. 19/331
Definition of a Simplex



Meshes in two and three dimensions are made of
polygons or polyhedra named elements.
The most simple amongst them, the simplices, are
those which have the minimal number of vertices.




                                         Mesh Quality – p. 19/331
Definition of a Simplex



Meshes in two and three dimensions are made of
polygons or polyhedra named elements.
The most simple amongst them, the simplices, are
those which have the minimal number of vertices.
The segment in one dimension.




                                         Mesh Quality – p. 19/331
Definition of a Simplex



Meshes in two and three dimensions are made of
polygons or polyhedra named elements.
The most simple amongst them, the simplices, are
those which have the minimal number of vertices.
The segment in one dimension.
The triangle in two dimensions.




                                         Mesh Quality – p. 19/331
Definition of a Simplex



Meshes in two and three dimensions are made of
polygons or polyhedra named elements.
The most simple amongst them, the simplices, are
those which have the minimal number of vertices.
The segment in one dimension.
The triangle in two dimensions.
The tetrahedron in three dimensions.




                                         Mesh Quality – p. 19/331
Definition of a Simplex



Meshes in two and three dimensions are made of
polygons or polyhedra named elements.
The most simple amongst them, the simplices, are
those which have the minimal number of vertices.
The segment in one dimension.
The triangle in two dimensions.
The tetrahedron in three dimensions.
The hypertetrahedron in four dimensions.



                                           Mesh Quality – p. 19/331
Definition of a Simplex



Meshes in two and three dimensions are made of
polygons or polyhedra named elements.
The most simple amongst them, the simplices, are
those which have the minimal number of vertices.
The segment in one dimension.
The triangle in two dimensions.
The tetrahedron in three dimensions.
The hypertetrahedron in four dimensions.
Quadrilaterals, pyramids, prisms, hexahedra and other
such aliens are named non-simplicial elements.

                                           Mesh Quality – p. 19/331
Definition of a d-Simplex in Rd


Let d + 1 points Pj = (p1j , p2j , . . . , pdj ) ∈ Rd , 1 ≤ j ≤ d + 1,
not in the same hyperplane, id est, such that the matrix of
order d + 1,
                                                     
                       p11 p12 · · · p1,d+1
                     p21 p22 · · · p2,d+1 
                                                     
                     .                           . 
               A= .    .
                               . ..
                               .
                               .           .      . 
                                                  . 
                    
                     pd1 pd2 · · · pd,d+1 
                        1      1 ···             1

be invertible. The convex hull of the points Pj is named the
d-simplex of points Pj .

                                                          Mesh Quality – p. 20/331
A Simplex Generates Rd


Any point X ∈ Rd , with Cartesian coordinates (xi )d , is
                                                   i=1
characterized by the d + 1 scalars λj = λj (X) defined as
solution of the linear system
                 d+1
              
              
              
                       pij λj = xi for 1 ≤ i ≤ d,
              
                  j=1
                  d+1
              
              
              
                       λj = 1,
              
                  j=1


whose matrix is A.



                                                     Mesh Quality – p. 21/331
What to Retain


In two dimensions, the simplex is a triangle.




                                            Mesh Quality – p. 22/331
What to Retain


In two dimensions, the simplex is a triangle.
In three dimensions, the simplex is a tetrahedron.




                                            Mesh Quality – p. 22/331
What to Retain


In two dimensions, the simplex is a triangle.
In three dimensions, the simplex is a tetrahedron.
The d + 1 vertices of a simplex in Rd give d vectors that
form a base of Rd .




                                            Mesh Quality – p. 22/331
What to Retain


In two dimensions, the simplex is a triangle.
In three dimensions, the simplex is a tetrahedron.
The d + 1 vertices of a simplex in Rd give d vectors that
form a base of Rd .
The coordinates λj (X) of a point X ∈ Rd in the base
generated by the simplex are the barycentric
coordinates.




                                            Mesh Quality – p. 22/331
Table of Contents


1. Introduction             8. Non-Simplicial
2. Simplex Definition          Elements
3. Degeneracies of          9. Shape Quality
  Simplices                   Visualization
4. Shape Quality of         10. Shape Quality
  Simplices                   Equivalence
5. Formulae for Simplices   11. Mesh Quality and
6. Voronoi, Delaunay and      Optimization
  Riemann                   12. Size Quality of
7. Shape Quality and          Simplices
  Delaunay                  13. Universal Quality
                            14. Conclusions



                                          Mesh Quality – p. 23/331
Degeneracy of Simplices


A d-simplex made of d + 1 vertices Pj is degenerate if its
vertices are located in the same hyperplane, id est, if the
matrix A is not invertible.




                                                 Mesh Quality – p. 24/331
Degeneracy of Simplices


A d-simplex is degenerate if its d + 1 vertices do not
generate the space Rd .




                                             Mesh Quality – p. 25/331
Degeneracy of Simplices


A d-simplex is degenerate if its d + 1 vertices do not
generate the space Rd .
Such is the case if the d + 1 vertices are located in a
space of dimension lower than d.




                                             Mesh Quality – p. 25/331
Degeneracy of Simplices


A d-simplex is degenerate if its d + 1 vertices do not
generate the space Rd .
Such is the case if the d + 1 vertices are located in a
space of dimension lower than d.
A triangle is degenerate if its vertices are collinear or
collapsed.




                                               Mesh Quality – p. 25/331
Degeneracy of Simplices


A d-simplex is degenerate if its d + 1 vertices do not
generate the space Rd .
Such is the case if the d + 1 vertices are located in a
space of dimension lower than d.
A triangle is degenerate if its vertices are collinear or
collapsed.
A tetrahedron is degenerate if its vertices are coplanar,
collinear or collapsed.




                                               Mesh Quality – p. 25/331
Degeneracy of Simplices


A d-simplex is degenerate if its d + 1 vertices do not
generate the space Rd .
Such is the case if the d + 1 vertices are located in a
space of dimension lower than d.
A triangle is degenerate if its vertices are collinear or
collapsed.
A tetrahedron is degenerate if its vertices are coplanar,
collinear or collapsed.
Nota bene : Strictly speaking, accordingly to the
definition, a degenerate simplex is no longer a simplex.


                                               Mesh Quality – p. 25/331
Degeneracy Criterion


A d-simplex is degenerate if its matrix A is not
invertible. A matrix is not invertible if its determinant is
null.




                                                Mesh Quality – p. 26/331
Degeneracy Criterion


A d-simplex is degenerate if its matrix A is not
invertible. A matrix is not invertible if its determinant is
null.
The size of a simplex is its area in two dimensions and
its volume in three dimensions.




                                                Mesh Quality – p. 26/331
Degeneracy Criterion


A d-simplex is degenerate if its matrix A is not
invertible. A matrix is not invertible if its determinant is
null.
The size of a simplex is its area in two dimensions and
its volume in three dimensions.
The size of a d-simplex K made of d + 1 vertices Pj is
given by
                  size(K) = det(A)/d!.




                                                Mesh Quality – p. 26/331
Degeneracy Criterion


A d-simplex is degenerate if its matrix A is not
invertible. A matrix is not invertible if its determinant is
null.
The size of a simplex is its area in two dimensions and
its volume in three dimensions.
The size of a d-simplex K made of d + 1 vertices Pj is
given by
                  size(K) = det(A)/d!.

A triangle is degenerate if its area is null.



                                                Mesh Quality – p. 26/331
Degeneracy Criterion


A d-simplex is degenerate if its matrix A is not
invertible. A matrix is not invertible if its determinant is
null.
The size of a simplex is its area in two dimensions and
its volume in three dimensions.
The size of a d-simplex K made of d + 1 vertices Pj is
given by
                  size(K) = det(A)/d!.

A triangle is degenerate if its area is null.
A tetrahedron is degenerate if its volume is null.


                                                Mesh Quality – p. 26/331
Taxonomy of Degenerate Simplices


This taxonomy is based on the different possible
degenerate states of the simplices.




                                           Mesh Quality – p. 27/331
Taxonomy of Degenerate Simplices


This taxonomy is based on the different possible
degenerate states of the simplices.
There are three cases of degenerate triangles.




                                           Mesh Quality – p. 27/331
Taxonomy of Degenerate Simplices


This taxonomy is based on the different possible
degenerate states of the simplices.
There are three cases of degenerate triangles.
There are ten cases of degenerate tetrahedra.




                                           Mesh Quality – p. 27/331
Taxonomy of Degenerate Simplices


This taxonomy is based on the different possible
degenerate states of the simplices.
There are three cases of degenerate triangles.
There are ten cases of degenerate tetrahedra.
In this classification, the four symbols
  ,     ,     and       stand for vertices of multiplicity
simple, double, triple and quadruple respectively.




                                               Mesh Quality – p. 27/331
1 – The Cap



    Name            h −→ 0                h=0

                      C
                      h
     Cap        A            B     A      C       B

Degenerate edges : None
Radius of the smallest circumcircle : ∞




                                              Mesh Quality – p. 28/331
2 – The Needle


   Name            h −→ 0                 h=0

                  C
                 h
   Needle       A           B       A,C              B

Degenerate edges : AC
Radius of the smallest circumcircle : hmax /2




                                                Mesh Quality – p. 29/331
3 – The Big Crunch



         Name         h −→ 0        h=0

                       C
                       h h B
          Big         A h           A,B,C
        Crunch

Degenerate edges : All
Radius of the smallest circumcircle : 0
The Big Crunch is the theory opposite of the Big Bang.




                                            Mesh Quality – p. 30/331
Degeneracy of Tetrahedra


 There is one case of degeneracy resulting in four
collapsed vertices.
 There are five cases of degeneracy resulting in four
collinear vertices.
 There are four cases of degeneracy resulting in four
coplanar vertices.
                      D                 D

                                          d
          A            C     A     a    C
                                    b
                   B                  B c


                                              Mesh Quality – p. 31/331
1 – The Fin


   Name            h −→ 0             h=0

                        D
                    h                     D
               A            C     A                C
    Fin
                        B                     B

Degenerate edges : None
Degenerate faces : One cap
Radius of the smallest circumsphere : ∞



                                              Mesh Quality – p. 32/331
2 – The Cap



   Name            h −→ 0             h=0

                     D
    Cap        A      h       C   A       D C
                          B                B
Degenerate edges : None
Degenerate faces : None
Radius of the smallest circumsphere : ∞




                                          Mesh Quality – p. 33/331
3 – The Sliver


  Name             h −→ 0               h=0

                           D
                                h            C
  Sliver       A            C       A             D
                       B                 B
Degenerate edges : None
Degenerate faces : None
Radius of the smallest circumsphere : rABC or ∞




                                          Mesh Quality – p. 34/331
4 – The Wedge



   Name            h −→ 0              h=0

                        D
                            h                    C, D
  Wedge        A            C      A
                       B                     B
Degenerate edges : CD
Degenerate faces : Two needles
Radius of the smallest circumsphere : rABC




                                             Mesh Quality – p. 35/331
5 – The Crystal



  Name            h −→ 0                  h=0

                          D
              A          h
  Crystal          h          C    A B       D C
                    B
Degenerate edges : None
Degenerate faces : Four caps
Radius of the smallest circumsphere : ∞




                                           Mesh Quality – p. 36/331
6 – The Spindle



  Name            h −→ 0               h=0

                            D
              A     h               A B, D     C
 Spindle            h           C
                        B
Degenerate edges : BD
Degenerate faces : Two caps and two needles
Radius of the smallest circumsphere : ∞




                                         Mesh Quality – p. 37/331
7 – The Splitter



 Name             h −→ 0               h=0

                  D
                  h          C
 Splitter    A                    A    D       B, C
                        h
                         B
Degenerate edges : BC
Degenerate faces : Two caps and two needles
Radius of the smallest circumsphere : ∞




                                           Mesh Quality – p. 38/331
8 – The Slat


 Name             h −→ 0                  h=0

              D
              h             C
   Slat                    h       A, D         B, C
              A
                           B
Degenerate edges : AD and BC
Degenerate faces : Four needles
Radius of the smallest circumsphere : hmax /2




                                            Mesh Quality – p. 39/331
9 – The Needle


  Name            h −→ 0              h=0

                         D
                          h
                        h hC      A      B, C, D
  Needle      A
                         B
Degenerate edges : BC, CD and DB
Degenerate faces : Three needles and one Big Crunch
Radius of the smallest circumsphere : hmax /2




                                         Mesh Quality – p. 40/331
10 – The Big Crunch



        Name           h −→ 0         h=0

                            D
          Big        A hh   C
                             h
                          h         A, B, C, D
        Crunch         h Bh

Degenerate edges : All
Degenerate faces : Four Big Crunches
Radius of the smallest circumsphere : 0




                                            Mesh Quality – p. 41/331
What to Retain


A triangle is degenerate if its vertices are collinear or
collapsed, hence if its area is null.




                                               Mesh Quality – p. 42/331
What to Retain


A triangle is degenerate if its vertices are collinear or
collapsed, hence if its area is null.
There are three cases of degeneracy of triangles.




                                               Mesh Quality – p. 42/331
What to Retain


A triangle is degenerate if its vertices are collinear or
collapsed, hence if its area is null.
There are three cases of degeneracy of triangles.
A tetrahedron is degenerate if its vertices are coplanar,
collinear or collapsed, hence if its volume is null.




                                               Mesh Quality – p. 42/331
What to Retain


A triangle is degenerate if its vertices are collinear or
collapsed, hence if its area is null.
There are three cases of degeneracy of triangles.
A tetrahedron is degenerate if its vertices are coplanar,
collinear or collapsed, hence if its volume is null.
There are ten cases of degeneracy of tetrahedra.




                                               Mesh Quality – p. 42/331
Table of Contents


1. Introduction             8. Non-Simplicial
2. Simplex Definition          Elements
3. Degeneracies of          9. Shape Quality
  Simplices                   Visualization
4. Shape Quality of         10. Shape Quality
  Simplices                   Equivalence
5. Formulae for Simplices   11. Mesh Quality and
6. Voronoi, Delaunay and      Optimization
  Riemann                   12. Size Quality of
7. Shape Quality and          Simplices
  Delaunay                  13. Universal Quality
                            14. Conclusions



                                          Mesh Quality – p. 43/331
Shape Quality of Simplices



An usual method used to quantify the quality of a mesh
is through the quality of the elements of that mesh.




                                          Mesh Quality – p. 44/331
Shape Quality of Simplices



An usual method used to quantify the quality of a mesh
is through the quality of the elements of that mesh.
A criterion usually used to quantify the quality of an
element is the shape measure.




                                             Mesh Quality – p. 44/331
Shape Quality of Simplices



An usual method used to quantify the quality of a mesh
is through the quality of the elements of that mesh.
A criterion usually used to quantify the quality of an
element is the shape measure.
This section is a guided tour of the shape measures
used for simplices.




                                             Mesh Quality – p. 44/331
The Regular Simplex


Definition : An element is regular if it maximizes its measure for
          a given measure of its boundary.




                                                      Mesh Quality – p. 45/331
The Regular Simplex


Definition : An element is regular if it maximizes its measure for
          a given measure of its boundary.
           The equilateral triangle is regular because it maximizes
           its area for a given perimeter.




                                                       Mesh Quality – p. 45/331
The Regular Simplex


Definition : An element is regular if it maximizes its measure for
          a given measure of its boundary.
           The equilateral triangle is regular because it maximizes
           its area for a given perimeter.
           The equilateral tetrahedron is regular because it
           maximizes its volume for a given surface of its faces.




                                                       Mesh Quality – p. 45/331
Simplicial Shape Measure


Definition A : A simplicial shape measure is a
continuous function that evaluates the shape of a simplex.
It must be invariant under translation, rotation, reflection
and uniform scaling of the simplex. A shape measure is
called valid if it is maximal only for the regular simplex and
if it is minimal for all degenerate simplices. Simplicial
shape measures are scaled to the interval [0, 1], and are 1
for the regular simplex and 0 for a degenerate simplex.




                                                  Mesh Quality – p. 46/331
Remarks


The invariance under translation, rotation and
reflection means that the simplicial shape measures
must be independent of the coordinates system.




                                         Mesh Quality – p. 47/331
Remarks


The invariance under translation, rotation and
reflection means that the simplicial shape measures
must be independent of the coordinates system.
The invariance under a valid uniform scaling means
that the simplicial shape measures must be
dimensionless (independent of the unit system).




                                          Mesh Quality – p. 47/331
Remarks


The invariance under translation, rotation and
reflection means that the simplicial shape measures
must be independent of the coordinates system.
The invariance under a valid uniform scaling means
that the simplicial shape measures must be
dimensionless (independent of the unit system).
The continuity means that the simplicial shape
measures must change continuously in function of the
coordinates of the vertices of the simplex.




                                          Mesh Quality – p. 47/331
The Radius Ratio


The radius ratio of a simplex K is a shape measure defined
as ρ = d ρK /rK , where ρK and rK are the radius of the
incircle and circumcircle of K (insphere and circumsphere
in 3D), and where d is the dimension of space.

                      K
                 ρK


                          rK




                                              Mesh Quality – p. 48/331
The Mean Ratio


Let R(r1 , r2 , r3 [, r4 ]) be an equilateral simplex having the
same [area|volume] than the simplex K(P1 , P2 , P3 [, P4 ]). Let
N be the matrix of transformation from R to K, i.e.
Pi = N ri + b, 1 ≤ i ≤ [3|4], where b is a translation vector.
                 s                             y      K
                              K = NR + b
                R
                      r
                                            b
                                                       x




                                                   Mesh Quality – p. 49/331
The Mean Ratio


Then, the mean ratio η of the simplex K is the ratio of the
geometric mean over the algebraic means of the
eigenvalues λ1 , λ2 [,λ3 ] of the matrix N T N .
                    √                √
                    2 λ1 λ2
                   
                        2
                                     4 3 SK
           d       
                    λ +λ =                          in 2D,
       d
              λi   1         2
                                               2
                                     1≤i<j≤3 Lij
          i=1
                   
  η=             =
           d       
       1
              λi    3 √λ 1 λ 2 λ 3
                   
                         3
                                         12 3 9VK2
       d           
                   
         i=1        λ +λ +λ =                   L 2
                                                     in 3D.
                      1    2    3      1≤i<j≤4   ij




                                                  Mesh Quality – p. 50/331
The Condition Number


F ORMAGGIA and P EROTTO (2000) use the inverse of the
condition number of the matrix.
                          min λiλ1
                            i
                     κ=        = ,
                        max λi  λd
                            i

if the eigenvalues are sorted in increasing order.




                                                 Mesh Quality – p. 51/331
The Frobenius Norm


Freitag and Knupp (1999) use the Frobenius norm of the
matrix N = AW −1 to define a shape measure.

                    d                                 d
   κ=                                 =                               ,
          tr(N T N )tr((N T N )−1 )         d              d
                                                 λi             λ−1
                                                                 i
                                           i=1            i=1


where the λi are the eigenvalues of the tensor N T N .




                                                          Mesh Quality – p. 52/331
The Minimum of Solid Angles


The simplicial shape measure θmin based on the minimum
of solid angles of the d-simplex is defined by

                   θmin = α−1 min θi ,
                               1≤i≤d+1


The coefficient α is the value of each solid angle of the
regular d-simplex, given by α = π/3 in two dimensions
                 √
and α = 6 arcsin 3/3 − π in three dimensions.




                                                Mesh Quality – p. 53/331
The sin of θmin


From a numerical point of view, a less expensive simplicial
shape measure is the sin of the minimum solid angle. This
avoids the computation of the arcsin(·) function in the
computation of θi in 2D and θi in 3D.

                    σmin = β −1 min σi ,
                               1≤i≤d+1


where σi = sin(θi ) in 2D and σi = sin(θi /2) in 3D. β is the
value of σi for all solid angles of the regular simplex, given
                  √                                √
by β = sin(α) = 3/2 in 2D and β = sin(α/2) = 6/9 in 3D.




                                                  Mesh Quality – p. 54/331
Face Angles


We can define a shape measure based on the minimum of
the twelve angles of the four faces of a tetrahedron. This
angle is π/3 for the regular tetrahedron.
But this shape measure is not valid according to
Definition A because it is insensitive to degenerate
tetrahedra that do not have degenerate faces (the sliver
and the cap).




                                               Mesh Quality – p. 55/331
Dihedral Angles


The dihedral angle is the angle between the intersection of
two adjacent faces to an edge with the perpendicular plane
of the edge.


                                        Pj
                       ϕij


                    Pi
The minimum of the six dihedral angles ϕmin is used as a
shape measure.

                                               Mesh Quality – p. 56/331
Dihedral Angles



   αϕmin = min ϕij = min (π − arccos (nij1 · nij2 )) ,
            1≤i<j≤4      1≤i<j≤4


where nij1 and nij2 are the normal to the adjacent faces of
the edge Pi Pj , and where α = π − arccos(−1/3) is the
value of the six dihedral angles of the regular tetrahedron.
But this shape measure is not valid according to
Definition A. The smallest dihedral angles of the needle,
the spindle and the crystal can be as large as π/3.




                                                 Mesh Quality – p. 57/331
The Interpolation Error Coefficient


In finite element, the interpolation error of a function over
an element is bounded by a coefficient times the
semi-norm of the function. This coefficient is the
ratio DK /̺K where DK is the diameter of the element K
and ̺K is the roundness of the element K.
                       √ ρK
                       2 3
                                    in 2 D,
                              hmax
                 γ=
                       2√6 ρK in 3 D.
                      
                              hmax




                                                  Mesh Quality – p. 58/331
The Edge Ratio


Ratio of the smallest edge over the tallest.

                       r = hmin /hmax .

The edge ratio r is not a valid shape measure according to
Definition A because it does not vanish for some
degenerate simplices. In 2D, it can be as large as 1/2 for
                                     √
the cap. In 3D, it can be as large as 2/2 for the sliver, 1/2
           √
for the fin, 3/3 for the cap and 1/3 for the crystal.




                                                  Mesh Quality – p. 59/331
Other Shape Measure – 1


 hmax /rK , the ratio of the diameter of the tetrahedron
over the circumradius, in B AKER, (1989). This is not a
valid shape measure.
 hmin /rK , the ratio of the smallest edge of the
tetrahedron over the circumradius, in M ILLER et al
(1996). This is not a valid shape measure.
 VK /rK 3 , the ratio of the volume of the tetrahedron over
the circumradius, in M ARCUM et W EATHERILL, (1995).




                                                Mesh Quality – p. 60/331
Other Shape Measure – 2


     4   4      2 −3
 VK      i=1   Si    ,
                   the ratio of the volume of the
tetrahedron over the area of its faces, in D E C OUGNY et
al (1990). The evaluation of this shape measure, and its
validity, are a complex problem for tetrahedra that
degenerate in four collinear vertices.
                         −3
VK       1≤i<j≤4   Lij        , the ratio of the volume of the
tetrahedron over the average of its edges, in
DANNELONGUE and TANGUY (1991), Z AVATTIERI et al
(1996) and W EATHERILL et al (1993).




                                                       Mesh Quality – p. 61/331
Other Shape Measure – 3


                         2
   VK              Lij       − L12 L34 − L13 L24
         1≤i<j≤4                                          −3/2

                             −L14 L23 +             L2
                                                     ij
                                          1≤i<j≤4
the ratio of the volume of the tetrahedron over a sum, at
the power three halfs, of many terms homogeneous to the
square of edge lenghts, in B ERZINS (1998).




                                                            Mesh Quality – p. 62/331
Other Shape Measure – 4


                       −3
VK               L2
          1≤i<j≤4 ij        , the ratio of the volume of the
tetrahedron over the quadratic average of the six edges,
in G RAICHEN et al (1991).
 And so on... This list is surely not exhaustive.




                                                   Mesh Quality – p. 63/331
There Exists an Infinity of Shape
                                            Measures


If µ and ν are two valid shape measures, if c, d ∈ R+ , then
    µc ,
    c(µ−1)/µ with c > 1,
    αµc + (1 − α)ν d with α ∈ [0, 1],
    µc ν d
are also valid simplicial shape measures.




                                                 Mesh Quality – p. 64/331
What to Retain


The regular simplex is the equilateral one, ie, where all
its edges have the same length.




                                             Mesh Quality – p. 65/331
What to Retain


The regular simplex is the equilateral one, ie, where all
its edges have the same length.
A shape measures evaluates the ratio to equilaterality.




                                             Mesh Quality – p. 65/331
What to Retain


The regular simplex is the equilateral one, ie, where all
its edges have the same length.
A shape measures evaluates the ratio to equilaterality.
A non valid shape measure does not vanish for all
degenerate simplices.




                                             Mesh Quality – p. 65/331
What to Retain


The regular simplex is the equilateral one, ie, where all
its edges have the same length.
A shape measures evaluates the ratio to equilaterality.
A non valid shape measure does not vanish for all
degenerate simplices.
There exists an infinity of valid shape measures.




                                             Mesh Quality – p. 65/331
What to Retain


The regular simplex is the equilateral one, ie, where all
its edges have the same length.
A shape measures evaluates the ratio to equilaterality.
A non valid shape measure does not vanish for all
degenerate simplices.
There exists an infinity of valid shape measures.
The goal of research is not to find an other one way
better than the other ones.




                                             Mesh Quality – p. 65/331
Table of Contents


1. Introduction             8. Non-Simplicial
2. Simplex Definition          Elements
3. Degeneracies of          9. Shape Quality
  Simplices                   Visualization
4. Shape Quality of         10. Shape Quality
  Simplices                   Equivalence
5. Formulae for Simplices   11. Mesh Quality and
6. Voronoi, Delaunay and      Optimization
  Riemann                   12. Size Quality of
7. Shape Quality and          Simplices
  Delaunay                  13. Universal Quality
                            14. Conclusions



                                          Mesh Quality – p. 66/331
Formulae for the Triangle


A triangle is completely defined by the knowledge of the
length of its three edges.

Quantities such that inradius, circumradius, angles, area,
etc, can be written in function of the edge lengths of the
triangle.

Let K be a non degenerate triangle of vertices P1 , P2
and P3 . The lengths of the edges Pi Pj of K are
denoted Lij = Pj − Pi , 1 ≤ i < j ≤ 3.




                                                Mesh Quality – p. 67/331
The Half-Perimeter


The half-perimeter pK is given by

                      (L12 + L13 + L23 )
                 pK =                    .
                              2




                                             Mesh Quality – p. 68/331
Heron’s Formula


The area SK of a triangle can also be written in function of
the edge lengths with Heron’s formula :
          2
         SK = pK (pK − L12 )(pK − L13 )(pK − L23 ).




                                                 Mesh Quality – p. 69/331
Radius of the Incircle


The radius ρK of the incircle of the triangle K is given by

                              SK
                         ρK =    .
                              pK




                                                  Mesh Quality – p. 70/331
Radius of the Circumscribed Circle


The radius rK of the circumcircle of the triangle K is given
by
                           L12 L13 L23
                     rK =              .
                              4SK




                                                 Mesh Quality – p. 71/331
Element Diameter


The diameter of an element is the biggest Euclidean
distance between two points of an element. For a triangle,
this is also the length of the biggest edge hmax

                hmax = max(L12 , L13 , L23 ),

The length of the smallest edge is denoted hmin

                 hmin = min(L12 , L13 , L23 ).




                                                  Mesh Quality – p. 72/331
Solid Angle


The angle θi at vertex Pi of triangle K is the arc length
obtained by projecting the edge of the triangle opposite
to Pi on a unitary circle centerered at Pi . The angle can be
written in function of the edge lengths as
                                                  −1
           θi = arcsin 2SK              Lij Lik        .
                                j,k=i
                              1≤j<k≤3




                                                       Mesh Quality – p. 73/331
Formulae for the Tetrahedron


A tetrahedron is completely defined by the knowledge of
the length of its six edges.

Quantities such that inradius, circumradius, angles,
volume, etc, can be written in function of the edge lengths
of the tetrahedron.




                                                Mesh Quality – p. 74/331
Formulae for the Tetrahedron


Let K be a non degenerate tetrahedron of vertices P1 , P2 ,
P3 and P4 . The lengths of the edges Pi Pj of K are denoted
Lij = Pj − Pi , 1 ≤ i < j ≤ 4. The area of the four faces of
the tetrahedron, △P2 P3 P4 , △P1 P3 P4 , △P1 P2 P4
and △P1 P2 P3 , are denoted by S1 , S2 , S3 and S4 . Finally, VK
is the volume of the tetrahedron K.




                                                   Mesh Quality – p. 75/331
3D “Heron’s” Formula


Let a, b, c, e, f and g be the length of the six edges of the
tetrahedron such that the edges a, b and c are connected
to the same vertex, and such that e is the opposite edge of
a, f is opposite of b and g is the opposite of c. The volume
VK is then
         2
     144VK = 4a2 b2 c2
           + (b2 + c2 − e2 ) (c2 + a2 − f 2 ) (a2 + b2 − g 2 )
                                2                     2
           − a2 (b2 + c2 − e2 ) − b2 (c2 + a2 − f 2 )
                                2
           − c2 (a2 + b2 − g 2 ) .




                                                      Mesh Quality – p. 76/331
Radius of the Insphere


The radius ρK of the insphere of the tetrahedron K is given
by
                              3VK
                  ρK =                    .
                        S1 + S2 + S3 + S4




                                               Mesh Quality – p. 77/331
Radius of the Circumsphere


The radius rK of the circumsphere of the tetrahedron K is
given by

             (a + b + c)(a + b − c)(a + c − b)(b + c − a)
    rK =                                                  .
                               24VK
where a = L12 L34 , b = L13 L24 and c = L14 L23 are the
product of the length of the opposite edges of K (two
edges are opposite if they do not share a vertex.




                                                   Mesh Quality – p. 78/331
Element Diameter


The diameter of an element is the biggest Euclidean
distance between two points of an element. For a
tetrahedron, this is also the length of the biggest edge hmax

          hmax = max(L12 , L13 , L14 , L23 , L24 , L34 ),

The length of the smallest edge is denoted hmin

           hmin = min(L12 , L13 , L14 , L23 , L24 , L34 ).




                                                         Mesh Quality – p. 79/331
Solid Angle


The solid angle θi at vertex Pi of the tetrahedron K, is the
area of the spherical sector obtained by projecting the face
of the tetrahedron opposite to Pi on a unitary sphere
centerered at Pi .
                                     P4



                     P1     θ1          P3


                                 P2


                                                Mesh Quality – p. 80/331
Solid angle


L IU and J OE (1994) gave a formula to compute the solid
angle in function of edge lengths :
                                                           −1/2
   θi = 2 arcsin 12VK             (Lij + Lik )2 − L2
                                                   jk              .
                          j,k=i
                        1≤j<k≤4




                                                        Mesh Quality – p. 81/331
Table of Contents

1. Introduction        8. Non-Simplicial
2. Simplex Definition     Elements
3. Degeneracies of     9. Shape Quality
  Simplices              Visualization
4. Shape Quality of    10. Shape Quality
  Simplices              Equivalence
5. Formulae for Sim-   11. Mesh Quality and
  plices                 Optimization
6. Voronoi, Delaunay   12. Size Quality of
  and Riemann            Simplices
7. Shape Quality and   13. Universal Quality
  Delaunay             14. Conclusions
                                   Mesh Quality – p. 82/331
Which Is the Most Beautiful Triangle ?




                            Mesh Quality – p. 83/331
Which Is the Most Beautiful Triangle ?




        A




                            Mesh Quality – p. 83/331
Which Is the Most Beautiful Triangle ?




        A                B




                             Mesh Quality – p. 83/331
If You Chose the Triangle A...




                             Mesh Quality – p. 84/331
If You Chose the Triangle A...




         A
   You are wrong !


                             Mesh Quality – p. 84/331
If You Chose the Triangle B...




                             Mesh Quality – p. 85/331
If You Chose the Triangle B...




                            B
                   You are wrong again !


                                Mesh Quality – p. 85/331
Which Is the Most Beautiful Triangle ?




          A               B
None of these answers !


                              Mesh Quality – p. 86/331
Which Is the Most Beautiful Woman ?




                          Mesh Quality – p. 87/331
Which Is the Most Beautiful Woman ?




       A


                          Mesh Quality – p. 87/331
Which Is the Most Beautiful Woman ?




       A                B


                            Mesh Quality – p. 87/331
You Probably chose...




                        Mesh Quality – p. 88/331
You Probably chose...




           A            B
Woman A.
                            Mesh Quality – p. 88/331
And if One Asked these Gentlemen...




                           Mesh Quality – p. 89/331
And if One Asked these Gentlemen...




                           Mesh Quality – p. 89/331
These Gentlemen Would Choose...




                          Mesh Quality – p. 90/331
These Gentlemen Would Choose...




           A           B
Woman B.
                           Mesh Quality – p. 90/331
Which Is the Most Beautiful Woman...

There is no absolute answer because the
question is incomplete.
One did not specify who was going to judge the
candidates, which was the scale of evaluation,
which were the measurements used, etc.




                                      Mesh Quality – p. 91/331
Which Is the Most Beautiful Triangle ?




                            Mesh Quality – p. 92/331
Which Is the Most Beautiful Triangle ?




        A                B



                             Mesh Quality – p. 92/331
Which Is the Most Beautiful Triangle ?




          A                          B
The question is incomplete : It misses a way of
measuring the quality of a triangle.

                                       Mesh Quality – p. 92/331
Voronoi Diagram

           Georgy Fedoseevich VORO -
           NOÏ . April 28, 1868, Ukraine
           – November 20, 1908, War-
           saw. Nouvelles applications
           des paramètres continus à
           la théorie des formes qua-
           dratiques. Recherches sur
           les parallélloèdes primitifs.
           Journal Reine Angew. Math,
           Vol 134, 1908.


                                Mesh Quality – p. 93/331
The Perpendicular Bisector

                         Let S1 and S2 be two
                         vertices in R2 . The
                         perpendicular       bisec-
 d(P, S1 )   P           tor M (S1 , S2 ) is the
 S1
               d(P, S2 ) locus of points equi-
                         distant to S1 and S2 .
                  S2 M (S1 , S2 ) = {P ∈
         M               R2 | d(P, S1 ) = d(P, S2 )},
                         where d(·, ·) is the Eucli-
                         dean distance between
                         two points of space.

                                          Mesh Quality – p. 94/331
A Cloud of Vertices

Let S = {Si }i=1,...,N be a cloud of N vertices.

                      S2        S11
                S9                     S10
           S5        S6    S4     S8
                S1
                          S7    S12      S3




                                              Mesh Quality – p. 95/331
The Voronoi Cell

Definition : The Voronoi cell C(Si ) associated to
the vertex Si is the locus of points of space which
is closer to Si than any other vertex :
  C(Si ) = {P ∈ R2 | d(P, Si ) ≤ d(P, Sj ), ∀j = i}.


                              C(Si )
                         Si



                                            Mesh Quality – p. 96/331
The Voronoi Diagram

The set of Voronoi cells associated with all the
vertices of the cloud of vertices is called the
Voronoi diagram.




                                         Mesh Quality – p. 97/331
Properties of the Voronoi Diagram

 The Voronoi cells are polygons in 2D,
 polyhedra in 3D and N -polytopes in N D.
 The Voronoi cells are convex.
 The Voronoi cells cover space without
 overlapping.




                                      Mesh Quality – p. 98/331
What to Retain

The Voronoi diagrams are partitions of space
into cells based on the concept of distance.




                                    Mesh Quality – p. 99/331
Delaunay Triangulation

            Boris Nikolaevich D ELONE or
            D ELAUNAY. 15 mars 1890,
            Saint Petersbourg — 1980.
            Sur la sphère vide. À la mé-
            moire de Georges Voronoi,
            Bulletin of the Academy of
            Sciences of the USSR, Vol. 7,
            pp. 793–800, 1934.




                                Mesh Quality – p. 100/331
Triangulation of a cloud of Points

The same cloud of points can be triangulated in
many different fashions.




                      ...



                                      Mesh Quality – p. 101/331
Triangulation of a Cloud of Points
                 ...




                 ...




                            Mesh Quality – p. 102/331
Triangulation of a Cloud of Points
                 ...




                 ...




                            Mesh Quality – p. 103/331
Delaunay Triangulation

Among all these fashions, there is one (or maybe
many) triangulation of the convex hull of the point
cloud that is said to be a Delaunay Triangulation.




                                        Mesh Quality – p. 104/331
Empty Sphere Criterion of Delaunay

Empty sphere criterion : A simplex K satisfies
the empty sphere criterion if the open
circumscribed ball of the simplex K is empty (ie,
does not contain any other vertex of the
triangulation).

                  K


                               K


                                       Mesh Quality – p. 105/331
Violation of the Empty Sphere Criterio

A simplex K does not satisfy the empty sphere
criterion if the opened circumscribed ball of
simplex K is not empty (ie, it contains at least
one vertex of the triangulation).
                              K
                  K




                                       Mesh Quality – p. 106/331
Delaunay Triangulation

Delaunay Triangulation : If all the simplices K
of a triangulation T satisfy the empty sphere
criterion, then the triangulation is said to be a
Delaunay triangulation.




                                        Mesh Quality – p. 107/331
Delaunay Algorithm

  The     circumscri-
 bed sphere of a
 simplex has to be      S3
 computed.
                                                 S2
  This amounts to                     ρout
 computing the cen-               C
 ter of a simplex.
  The center is the
 point at equal dis-
 tance to all the
 vertices of the sim-        S1
 plex.
                                         Mesh Quality – p. 108/331
Delaunay Algorithm

How can we know if a point P violates the empty
sphere criterion for a simplex K ?
   The center C and the radius ρ of the
  circumscribed sphere of the simplex K has to
  be computed.
   The distance d between the point P and the
  center C has to be computed.
   If the distance d is greater than the radius ρ,
  the point P is not in the circumscribed sphere
  of the simplex K.


                                        Mesh Quality – p. 109/331
What to Retain

The Voronoi diagram of a cloud of points is a
partition of space into cells based on the
notion of distance.

A Delaunay triangulation of a cloud of points
is a triangulation based on the notion of
distance.




                                    Mesh Quality – p. 110/331
Duality Delaunay-Voronoï

The Voronoï diagram is the dual of the Delaunay
triangulation and vice versa.




                                      Mesh Quality – p. 111/331
Voronoï and Delaunay in Nature

Voronoï diagrams and Delaunay triangulations
are not just a mathematician’s whim, they
represent structures that can be found in nature.




                                       Mesh Quality – p. 112/331
Voronoï and Delaunay In Nature




                          Mesh Quality – p. 113/331
A Turtle




           Mesh Quality – p. 114/331
A Pineapple




              Mesh Quality – p. 115/331
The Devil’s Tower




                    Mesh Quality – p. 116/331
Dry Mud




          Mesh Quality – p. 117/331
Bee Cells




            Mesh Quality – p. 118/331
Dragonfly Wings




                 Mesh Quality – p. 119/331
Pop Corn




           Mesh Quality – p. 120/331
Fly Eyes




           Mesh Quality – p. 121/331
Carbon Nanotubes




                   Mesh Quality – p. 122/331
Soap Bubbles




               Mesh Quality – p. 123/331
A Geodesic Dome




                  Mesh Quality – p. 124/331
Biosphère de Montréal




                        Mesh Quality – p. 125/331
Streets of Paris




                   Mesh Quality – p. 126/331
Roads in France




                  Mesh Quality – p. 127/331
Roads in France




                  Mesh Quality – p. 128/331
Where Is this Guy Going ? ! !

   A simplicial shape measure is an evaluation
   of the ratio to equilarity.




                                     Mesh Quality – p. 129/331
Where Is this Guy Going ? ! !

   A simplicial shape measure is an evaluation
   of the ratio to equilarity.
   The Voronoï diagram of a cloud of points is a
   partition of space into cells based on the
   notion of distance.




                                      Mesh Quality – p. 129/331
Where Is this Guy Going ? ! !

   A simplicial shape measure is an evaluation
   of the ratio to equilarity.
   The Voronoï diagram of a cloud of points is a
   partition of space into cells based on the
   notion of distance.
   A Delaunay triangulation of a cloud of points
   is a triangulation based on the notion of
   distance.



                                       Mesh Quality – p. 129/331
Where Is this Guy Going ? ! !

   A simplicial shape measure is an evaluation
   of the ratio to equilarity.
   The Voronoï diagram of a cloud of points is a
   partition of space into cells based on the
   notion of distance.
   A Delaunay triangulation of a cloud of points
   is a triangulation based on the notion of
   distance.
   The notion of distance can be generalized.


                                       Mesh Quality – p. 129/331
Where Is this Guy Going ? ! !

   A simplicial shape measure is an evaluation
   of the ratio to equilarity.
   The Voronoï diagram of a cloud of points is a
   partition of space into cells based on the
   notion of distance.
   A Delaunay triangulation of a cloud of points
   is a triangulation based on the notion of
   distance.
   The notion of distance can be generalized.
   The notions of shape measure, of Voronoï
   diagram and of Delaunay triangulation Quality –be
                                      Mesh
                                           can p. 129/331
Nikolai Ivanovich Lobachevsky


             N IKOLAI     I VANOVICH
             LOBACHEVSKY,          1
             décembre 1792, Nizhny
             Novgorod — 24 février
             1856, Kazan.




                              Mesh Quality – p. 130/331
János Bolyai

               J ÁNOS BOLYAI, 15 dé-
               cembre 1802 à Kolozsvár,
               Empire Austrichien (Cluj,
               Roumanie) — 27 janvier
               1860 à Marosvásárhely,
               Empire Austrichien (Tirgu-
               Mures, Roumanie).




                                 Mesh Quality – p. 131/331
Bernhard RIEMANN

         G EORG F RIEDRICH B ERN -
         HARD RIEMANN, 7 sep-
         tembre 1826, Hanovre — 20
         juillet 1866, Selasca. Über die
         Hypothesen welche der Geo-
         metrie zu Grunde liegen. 10
         juin 1854.




                               Mesh Quality – p. 132/331
Non Euclidean Geometry

Riemann has generalized Euclidean geometry in
the plane to Riemannian geometry on a surface.
He has defined the distance between two points
on a surface as the length of the shortest path
between these two points (geodesic).
He has introduced the Riemannian metric that
defines the curvature of space.




                                      Mesh Quality – p. 133/331
The Metric in the Merriam-Webster




                          Mesh Quality – p. 134/331
Definition of a Metric

If S is any set, then the function
                    d : S×S → I
                              R
is called a metric on S if it satisfies
(i) d(x, y) ≥ 0 for all x, y in S ;
(ii) d(x, y) = 0 if and only if x = y ;
(iii) d(x, y) = d(y, x) for all x, y in S ;
(iv) d(x, y) ≤ d(x, z) + d(z, y) for all x, y, z in S.




                                             Mesh Quality – p. 135/331
The Euclidean Distance is a Metric

In the previous definition of a metric, let the set S
be I 2 , the function
    R
       d : I 2 ×I 2 → I
           R R        R
     x1     x2
          ×            →   (x2 − x1 )2 + (y2 − y1 )2
     y1     y2

is a metric on I 2 .
               R




                                           Mesh Quality – p. 136/331
Metric Space




               Mesh Quality – p. 137/331
The Scalar Product is a Metric

Let a vectorial space with its scalar product ·, · .
Then the norm of the scalar product of the
difference of two elements of the vectorial space
is a metric.
         d(A, B) =      B−A ,
                                       1/2
                   =   B − A, B − A          ,
                       − − 1/2
                        → →
                   =   AB, AB    ,
                         − T−
                          → →
                   =     AB AB.

                                             Mesh Quality – p. 138/331
The Scalar Product is a Metric

If the vectorial space is I 2 , then the norm of the
                          R
                                − →
scalar product of the vector AB is the Euclidean
distance.

                                   1/2       − T−
                                              → →
  d(A, B) =     B − A, B − A             =   AB AB,
                               T
                    xB − xA          xB − xA
            =                                 ,
                    y B − yA         y B − yA

            =     (xB − xA )2 + (yB − yA )2 .

                                              Mesh Quality – p. 139/331
Metric Tensor

A metric tensor M is a symmetric positive
definite matrix

                   m11 m12
           M=                 in 2D,
                   m12 m22
                              
              m11 m12 m13
                         
        M =  m12 m22 m23  in 3D.
              m13 m23 m33


                                       Mesh Quality – p. 140/331
Metric Length
               −→
The length LM (AB) of an edge between vertices
A and B in the metric M is given by
             −→       − − 1/2
                       → →
         LM (AB) =    AB, AB M ,
                      −→    − 1/2
                             →
                  =   AB, M AB    ,
                       − T −
                        →     →
                  =    AB M AB.




                                      Mesh Quality – p. 141/331
Euclidean Length with M = I

    −→      −→    −→    1/2       − T −
                                   →    →
LM (AB) =   AB, M AB          =   AB M AB,
                              T
                xB − xA           1 0         xB − xA
       =
                y B − yA          0 1         y B − yA
    −→
LE (AB) =    (xB − xA )2 + (yB − yA )2 .




                                        Mesh Quality – p. 142/331
αβ
Metric Length with M =                  βγ

    −→      −→    −→    1/2       − T −
                                   →    →
LM (AB) =   AB, M AB          =   AB M AB,
                              T
                xB − xA           α β           xB − xA
       =
                y B − yA          β γ           y B − yA
    −→
LE (AB) =   α(xB − xA )2 + 2β(xB − xA )(yB − yA )
                                   2 1/2
                    +γ(yB − yA )           .



                                         Mesh Quality – p. 143/331
Length in a Variable Metric

In the general sense, the metric tensor M is not
constant but varies continuously for every point
of space. The length of a parameterized curve
γ(t) = {(x(t), y(t), z(t)) , t ∈ [0, 1]} is evaluated in
the metric
                     1
     LM (γ) =            (γ ′ (t))T M (γ(t)) γ ′ (t) dt,
                 0

where γ(t) is a point of the curve and γ ′ (t) is the
tangent vector of the curve at that point. LM (γ) is
always bigger or equal to the geodesic between
the end points of the curve.
                                                 Mesh Quality – p. 144/331
Area and Volume in a Metric

Area of the triangle K in a metric M :

           AM (K) =        det(M) dA.
                      K


Volume of the tetrahedron K in a metric M :

           VM (K) =        det(M) dV.
                      K




                                         Mesh Quality – p. 145/331
Metric and Delaunay Mesh




                           Mesh Quality – p. 146/331
Which is the Best Triangle ?




            A                         B
The question is incomplete. The way to measure
the quality of the triangle is missing.

                                     Mesh Quality – p. 147/331
Which is the Best Triangle ?




        A                B



                               Mesh Quality – p. 148/331
Which is the Best Triangle ?




        A                B



                               Mesh Quality – p. 149/331
Example of an Adapted Mesh




Adapted mesh and solution for a transonic
visquous compressible flow with Mach 0.85 and
Reynolds = 5 000.
                                    Mesh Quality – p. 150/331
Zoom on Boundary Layer–Shock




                        Mesh Quality – p. 151/331
What to Retain

  Beauty, quality and shape are relative
  notions.




                                   Mesh Quality – p. 152/331
What to Retain

  Beauty, quality and shape are relative
  notions.
  We first need to define what we want in order
  to evaluate what we obtained.




                                   Mesh Quality – p. 152/331
What to Retain

  Beauty, quality and shape are relative
  notions.
  We first need to define what we want in order
  to evaluate what we obtained.
  “What we want” is written in the form of metric
  tensors.




                                      Mesh Quality – p. 152/331
What to Retain

  Beauty, quality and shape are relative
  notions.
  We first need to define what we want in order
  to evaluate what we obtained.
  “What we want” is written in the form of metric
  tensors.
  A shape measure is a measure of the
  equilarity of a simplex in this metric.



                                      Mesh Quality – p. 152/331
Shape Measure in a Metric

First method (constant metric)


For a simplex K, evaluate the metric tensor at
several points (Gaussian points) and find an
averaged metric tensor.
Take this averaged metric tensor as constant
over the whole simplex and evaluate the shape
measure using this metric.



                                       Mesh Quality – p. 153/331
Shape Measure in a Metric

Second method (constant metric)


For a simplex K, evaluate the metric tensor at
one point (Gaussian point) and take the metric
as constant over the whole simplex. Evaluate the
shape measure using this metric.
Repeat this operation at several points and
average the shape measures.
This is what is done at INRIA.

                                       Mesh Quality – p. 154/331
Shape Measure in a Metric

Third methode (variable metric)


Express the shape measure as a fonction of
edge lengths only.
Evaluate the length of the edges in the metric
and compute the shape measure with these
lengths.
This is what is done in OORT.


                                       Mesh Quality – p. 155/331
Shape Measure in a Metric

Fourth method (variable metric)


Express the shape measure in function of the
length of the edges, the area and the volumes.
Evaluate the lengths, the area and the volume in
the metric.




                                      Mesh Quality – p. 156/331
Shape Measure in a Metric

Fifth method (variable metric)
Know how to evaluate quantities such as the
radius of the inscribed circle, of the
circumscribed circle, the solid angle, etc, in a
metric.
In the general sense, the triangular inequality is
not verified in a variable metric. Neither is the
sum of the angles equal to 180 degrees, etc.
The evaluation of a shape measure in a variable
metric in all its generality is an opened problem.
For the moment, it is approximated.
                                         Mesh Quality – p. 157/331
Table of Contents

1. Introduction        8. Non-Simplicial
2. Simplex Definition     Elements
3. Degeneracies of     9. Shape Quality
  Simplices              Visualization
4. Shape Quality of    10. Shape Quality
  Simplices              Equivalence
5. Formulae for Sim-   11. Mesh Quality and
  plices                 Optimization
6. Voronoi, Delaunay   12. Size Quality of
  and Riemann            Simplices
7. Shape Quality and   13. Universal Quality
  Delaunay             14. Conclusions
                                  Mesh Quality – p. 158/331
Shape Measures and Delaunay Critero
Delaunay meshes have several smoothness
properties.
  The Delaunay mesh minimizes the maximum value of
 all the element circumsphere radii.
  When the circumsphere center of all simplices of a
 mesh lie in their respective simplex, then the mesh is a
 Delaunay mesh.
  In a Delaunay mesh, the sum of all squared edge
 lengths weighted by the volume of elements sharing that
 edge is minimal.



                                             Mesh Quality – p. 159/331
3D-Delaunay Mesh and Degeneracy

In three dimensions, it is well known that
Delaunay meshes can include slivers which are
degenerate elements.

Why ?

How to avoid them ?




                                     Mesh Quality – p. 160/331
Empty Sphere Criterion of Delaunay




The empty sphere criterion of Delaunay is not a
shape measure, but it can be used like a shape
measure in an edge swapping algorithm.



                                      Mesh Quality – p. 161/331
Edge Swapping and θmin Shape Measu
During edge swapping, using the empty sphere
criterion (Delaunay criterion)
                        ⇐⇒
Using the θmin shape measure (maximize the
minimum of the angles).


                θ3             θ3 θ6
           θ1        θ2   θ1           θ5
           θ4        θ6
                θ5             θ2 θ4


                                            Mesh Quality – p. 162/331
What to Retain

  The empty sphere criterion of Delaunay is not
  a shape measure but it can be used as a
  shape measure.




                                     Mesh Quality – p. 163/331
What to Retain

  The empty sphere criterion of Delaunay is not
  a shape measure but it can be used as a
  shape measure.
  In two dimensions, in the edge swapping
  algorithm (Lawson’s method), the empty
  sphere criterion of Delaunay is equivalent to
  the θmin shape measure.




                                      Mesh Quality – p. 163/331
What to Retain

  The empty sphere criterion of Delaunay is not
  a shape measure but it can be used as a
  shape measure.
  In two dimensions, in the edge swapping
  algorithm (Lawson’s method), the empty
  sphere criterion of Delaunay is equivalent to
  the θmin shape measure.
  There is a multitude of valid shape measures,
  and thus a multitude of generalizations of the
  Delaunay mesh.

                                      Mesh Quality – p. 163/331
Delaunay and Circumscribed Sphere

As the circumscribed sphere of a tetrahedron
gets larger, there are more chances that another
vertex of the mesh happens to be in this sphere,
and the chances that this tetrahedron and the
mesh satisfy the Delaunay criterion get smaller.
As the circumscribed sphere of a tetrahedron
gets smaller, there are fewer chances that
another vertex of the mesh happens to be in this
sphere, and the chances that this tetrahedron
and the mesh satisfy the Delaunay criterion get
bigger.
                                      Mesh Quality – p. 164/331
Circumscribed Sphere of Infinite Radi

The tetrahedra that degenerate into a fin, into a
cap, into a crystal, into a spindle and into a
splitter
          D              D                     D
        h                           A          h
  A          C     A      h C           h        C
           B                B            B
                  D               D
          A    h                          C
               h       C     A h       h
                 B                      B
have a circumscribed sphere of infinite radius.


                                         Mesh Quality – p. 165/331
Circumscribed Sphere of Bounded Ra

The tetrahedra that degenerate into a sliver, into
a wedge, into a slat, into a needle and into a
Big Crunch
          D                 D
             h                 h    D
 A         C       A          C      h          C
                                    A           h
         B                  B                  B
                       D            D
                        h
                         C    A hh h C
                     hh            h
           A
                       B         h Bh
have a circumscribed sphere of bounded radius.


                                        Mesh Quality – p. 166/331
What to Retain

The empty sphere criterion of
Delaunay is not a valid shape
measure sensitive to all the possible
degeneracies of the tetrahedron.




                              Mesh Quality – p. 167/331
Circumscribed Sphere of Bounded Ra

Amongst the degenerate tetrahedra that have a
circumscribed sphere of bounded radius, the
wedge, the slat, the needle and the Big Crunch
can be eliminitated
                  D
                     h    D
          A         C     h           C
                          A          h
                  B                  B
                      D           D
                       h
                        C    A hh h C
                     hh          h
           A
                       B       h Bh
since they have several superimposed vertices.

                                    Mesh Quality – p. 168/331
The Sliver

And so, finally, we come to the sliver,
                  D
                     h               C
         A         C      A            D
                 B                B
a degenerate tetrahedron having disjoint vertices
and a bounded circumscribed sphere radius,
which makes it “Delaunay-admissible”.




                                       Mesh Quality – p. 169/331
Non-Convex Quadrilateral

It is forbidden to swap an edge of a non-convex
quadrilateral.
               S3           S3
                             T1
            T1 T2            S2
              S2             T2
       S1           S4 S1           S4
                       S3

                      T1                 S2
                                         T2
                S1          S4 S1             S4
                                              Mesh Quality – p. 170/331
Non-Convex Quadrilateral


       S3          Two adjacent triangles
                   forming a non-convex
                   quadrilateral necessa-
      T1 T2        rily satisfy the empty
        S2         sphere     criterion of
 S1           S4   Delaunay.




                                Mesh Quality – p. 171/331
Loss of the Convexity Property in 3D




                           Mesh Quality – p. 172/331
What to Retain

  The empty sphere criterion of Delaunay is
  more or less a simplicial shape measure.




                                    Mesh Quality – p. 173/331
What to Retain

  The empty sphere criterion of Delaunay is
  more or less a simplicial shape measure.
  The empty sphere criterion of Delaunay is not
  sensitive to all the possible degeneracies of
  the tetrahedron.




                                     Mesh Quality – p. 173/331
What to Retain

  The empty sphere criterion of Delaunay is
  more or less a simplicial shape measure.
  The empty sphere criterion of Delaunay is not
  sensitive to all the possible degeneracies of
  the tetrahedron.
  A valid shape measure, sensitive to all the
  possible degeneracies of the tetrahedron,
  used in an edge swapping and face swapping
  algorithm should lead to a mesh that is not a
  Delaunay mesh, but that is of better quality.

                                     Mesh Quality – p. 173/331
Table of Contents

1. Introduction        8. Non-Simplicial
2. Simplex Definition     Elements
3. Degeneracies of     9. Shape Quality
  Simplices              Visualization
4. Shape Quality of    10. Shape Quality
  Simplices              Equivalence
5. Formulae for Sim-   11. Mesh Quality and
  plices                 Optimization
6. Voronoi, Delaunay   12. Size Quality of
  and Riemann            Simplices
7. Shape Quality and   13. Universal Quality
  Delaunay             14. Conclusions
                                  Mesh Quality – p. 174/331
Non-Simplicial Elements

This section proposes a method to generalize
the notions of regularity, of degeneration and of
shape measure of simplices to non simplicial
elements ; i.e., to quadrilaterals in two
dimensions, to prisms and hexahedra in three
dimensions.




                                        Mesh Quality – p. 175/331
Non-Simplicial Elements

On Element Shape Measures for Mesh
Optimization
PAUL L ABBÉ , J ULIEN D OMPIERRE , F RANÇOIS
G UIBAULT AND R ICARDO C AMARERO
Presented at the 2nd Symposium on Trends in
Unstructured Mesh Generation, Fifth US National
Congress on Computational Mechanics, 4–6
august 1999 University of Colorado at Boulder.



                                     Mesh Quality – p. 176/331
Regularity Generalization

  An equilateral quadrilateral, ie that has four
  edges of same length, is not necessarily a
  square...




                                       Mesh Quality – p. 177/331
Regularity Generalization

  An equilateral quadrilateral, ie that has four
  edges of same length, is not necessarily a
  square...
  Définition : An element, be it simplicial or
  not, is regular if it maximizes its measure for a
  given measure of its boundary.




                                       Mesh Quality – p. 177/331
Regularity Generalization

  An equilateral quadrilateral, ie that has four
  edges of same length, is not necessarily a
  square...
  Définition : An element, be it simplicial or
  not, is regular if it maximizes its measure for a
  given measure of its boundary.
  The equilateral triangle is regular because it
  maximizes its area for a given perimiter.



                                       Mesh Quality – p. 177/331
Regularity Generalization

  An equilateral quadrilateral, ie that has four
  edges of same length, is not necessarily a
  square...
  Définition : An element, be it simplicial or
  not, is regular if it maximizes its measure for a
  given measure of its boundary.
  The equilateral triangle is regular because it
  maximizes its area for a given perimiter.
  The equilateral tetrahedron is regular
  because it maximizes its volume for a given
  surface of its faces.
                                       Mesh Quality – p. 177/331
Regular Non Simplicial Elements

  The regular quadrilateral is the square.
  The regular hexahedron is the cube.
  The regular prism is the ... regular prism ! Its
 two triangular faces are equilateral triangle
 whose edges measure a. Its three quadrilateral
 faces are rectangles that have a base of
                                     √
 length a and a height of length a/ 3.




                                       Mesh Quality – p. 178/331
Quality of Non Simplicial Elements

Proposed Extension : The shape measure of a
non simplicial element is given by the minimum
shape measure of the corner simplices
constructed from each vertex of the element and
of its neighbors.




                                     Mesh Quality – p. 179/331
Shape Measure of a Quadrilateral

The shape measure of a quadrilateral is the
minimum of the shape measure of its four corner
triangles formed by its four vertices.
       D C      D            C D C D C

     A    B   A    BA     B      BA




                                      Mesh Quality – p. 180/331
Shape Measure of a Prism

The shape measure of a prism is the minimum of
the shape measure of its six corner tetrahedron
formed by its six vertices.
                                         F
               D              E
        F               C       C         C
  D    E       A B A B A B
         C             F       F         F
   A B         D     E D      E D       E
                                          C
               A             B

                                     Mesh Quality – p. 181/331
Shape Measure of an Hexahedron

The shape measure of an hexahedron is the
minimum of its eight corner tetrahedron formed
by its eight vertices.
                                    G     H
              E            F
   H G            D           C D C       D C
E       F     AHB A       B HB A H
    D C                      G      G        G
A      B
              E      F E   F       F E
                                     C    D
              A           B
                                     Mesh Quality – p. 182/331
Shape of the Corner Simplex

  The corner simplices constructed for the non
 simplicial elements are not regular simplices.
  For the square, the four corner triangles are
 isosceles right-angled triangles.
  For the cube, the eight corner tetrahedra are
 isosceles right-angled tetrahedra.
  For the regular prism, the six corner
 tetrahedra are tetrahedron with an equilateral
 triangle of side a, √ a fourth perpendicular
                     and
 edge of length a/ 3.


                                      Mesh Quality – p. 183/331
Shape of the Corner Simplex

Each non simplicial shape measure has to be
normalized so as to be a shape measure equal
to unit value for regular non simplicial elements.
                ρ            η               θmin                     γ
                             √                                      √
                2             3                3                      3
 Square         √
              1+ 2           2                 4
                                                                     √
                                                                   1+ 2
                                              √         √           √
                 18√          1      2 arcsin(1/ 22+12 3)          3 √6
 Prism    √
              5(7+ 13)
                             √
                             3
                               2
                                                    √
                                         6 arcsin(1/ 3)−π         7+ 13
           √                 √                     √    √       √
                         2   3       2 arcsin((2− 2)/(2 3))
 Cube          3−1       3       2                  √
                                         6 arcsin(1/ 3)−π
                                                                    3−1

                                                     Mesh Quality – p. 184/331
Degenerate Non Simplicial Elements

Définition :A non simplicial element is
degenerate if at least one of its corner simplices
is degenerate.
If at least one of the corner simplices is more
than degenerate, meaning that it is inverted (of
negative norm), then the non simplicial element
is concave and is also considered degenerate.




                                        Mesh Quality – p. 185/331
Twisted Non Simplicial Elements

In three dimensions, the definition of the shape
measure of non simplicial elements has one
flaw : it is not sensitive to twisted elements.
                                  E
                       D              F
           E             C          C       C
  F       D      A B A B A B
            C           E          E      E
   A B           F     D F       D F      D
                                            C
                 A               B

                                       Mesh Quality – p. 186/331
Twist of Quadrilateral Faces

A critera used to measure the twist of a
quadrilateral face ABCD is to consider the
dihedral angle between the triangles ABC
and ACD on one hand, and between the
triangles ABD and BCD on the other hand.
If these dihedral angles are π, then the
quadrilateral face is a plane (not twisted). The
twist in the quadrilateral increases as the angles
differ from π.


                                        Mesh Quality – p. 187/331
Twist of Quadrilateral Faces

Definition :Given a valid simplicial shape
measure, the twist of a quadrilateral face is equal
to the value of the shape measure for the
tetrahedron constructed by the four vertices of
the quadrilateral face.
Thus, a plane face has no twist because the four vertices
form a degenerated tetrahedron and all valid shape
measures are null.
As a quadrilateral face is twisted, its vertices move away
from coplanarity, and the shape measure of the generated
tetrahedron gets larger.
                                              Mesh Quality – p. 188/331
What to Retain

The shape, the degeneration, the convexity,
the concavity and the torsion can be rewritten
as a function of simplices.
An advantage of this approach is that once that
the measurement and the shape measures for
the simplices are programmed, in Euclidean as
well as with a Riemannian metric, the extension
for non simplicial elements is direct.




                                      Mesh Quality – p. 189/331
Table of Contents
1. Introduction        8. Non-Simplicial
2. Simplex Definition     Elements
3. Degeneracies of     9. Shape Quality
  Simplices              Visualization
4. Shape Quality of    10. Shape Quality
  Simplices              Equivalence
5. Formulae for Sim-   11. Mesh Quality and
  plices                 Optimization
6. Voronoi, Delaunay   12. Size Quality of
  and Riemann            Simplices
7. Shape Quality and   13. Universal Quality
  Delaunay             14. Conclusions

                                    Mesh Quality – p. 190/331
Visualizing Shape Measures

                               1

                              0.5

                 QK (C)
       y                       2
                 C(x, y) x
A(0, 1/2)                           1



B(0, −1/2)   1                      y 0

                                          -1                               3

              0                                -2
                                                    0
                                                        1     x
                                                                  2



  Position of the three vertices A, B and C of the
  triangle K used to construct the contour plots of
  a shape measure.


                                                            Mesh Quality – p. 191/331
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality
Mesh Quality

More Related Content

Similar to Mesh Quality

MM4MPR Presentation
MM4MPR PresentationMM4MPR Presentation
MM4MPR Presentation
Thomas Arnold
 
Optimization of Heavy Vehicle Suspension System Using Composites
Optimization of Heavy Vehicle Suspension System Using CompositesOptimization of Heavy Vehicle Suspension System Using Composites
Optimization of Heavy Vehicle Suspension System Using Composites
IOSR Journals
 
Finite Element Analysis of PNC - Copy
Finite Element Analysis of PNC - CopyFinite Element Analysis of PNC - Copy
Finite Element Analysis of PNC - Copy
Raj Vaibhav
 
Leveling and Aligning
Leveling and AligningLeveling and Aligning
Leveling and Aligning
Pranshu Mathur
 
6796.optical fibres
6796.optical fibres6796.optical fibres
6796.optical fibres
Ankush Saini
 
buckling analysis of cantilever pultruded I-sections using 𝐴𝑁𝑆𝑌𝑆 ®
buckling analysis of cantilever pultruded I-sections using 𝐴𝑁𝑆𝑌𝑆 ®buckling analysis of cantilever pultruded I-sections using 𝐴𝑁𝑆𝑌𝑆 ®
buckling analysis of cantilever pultruded I-sections using 𝐴𝑁𝑆𝑌𝑆 ®
IJARIIE JOURNAL
 
Analysis of elliptical steel tubular section with frp
Analysis of elliptical steel tubular section with frpAnalysis of elliptical steel tubular section with frp
Analysis of elliptical steel tubular section with frp
IRJET Journal
 
Cq31401405
Cq31401405Cq31401405
Cq31401405
IJMER
 
Flexural behavioural study on rc beam with
Flexural behavioural study on rc beam withFlexural behavioural study on rc beam with
Flexural behavioural study on rc beam with
eSAT Publishing House
 
yarn manu. carding,blowroom
yarn manu. carding,blowroomyarn manu. carding,blowroom
yarn manu. carding,blowroom
Sumit Gwala
 
Effect of Diaphragms in Axially Loaded Timber Hollow.pptx
Effect of Diaphragms in Axially Loaded Timber Hollow.pptxEffect of Diaphragms in Axially Loaded Timber Hollow.pptx
Effect of Diaphragms in Axially Loaded Timber Hollow.pptx
DSK Presentation Hub
 
Michael Berkson Summer REU Extended Abstract
Michael Berkson Summer REU Extended AbstractMichael Berkson Summer REU Extended Abstract
Michael Berkson Summer REU Extended Abstract
Michael Berkson
 
Young’s modulus 2 and resistivity
Young’s modulus 2 and resistivityYoung’s modulus 2 and resistivity
Young’s modulus 2 and resistivity
HorsforthPhysics
 
Composite materials lecture
Composite materials lectureComposite materials lecture
Composite materials lecture
Padmanabhan Krishnan
 

Similar to Mesh Quality (14)

MM4MPR Presentation
MM4MPR PresentationMM4MPR Presentation
MM4MPR Presentation
 
Optimization of Heavy Vehicle Suspension System Using Composites
Optimization of Heavy Vehicle Suspension System Using CompositesOptimization of Heavy Vehicle Suspension System Using Composites
Optimization of Heavy Vehicle Suspension System Using Composites
 
Finite Element Analysis of PNC - Copy
Finite Element Analysis of PNC - CopyFinite Element Analysis of PNC - Copy
Finite Element Analysis of PNC - Copy
 
Leveling and Aligning
Leveling and AligningLeveling and Aligning
Leveling and Aligning
 
6796.optical fibres
6796.optical fibres6796.optical fibres
6796.optical fibres
 
buckling analysis of cantilever pultruded I-sections using 𝐴𝑁𝑆𝑌𝑆 ®
buckling analysis of cantilever pultruded I-sections using 𝐴𝑁𝑆𝑌𝑆 ®buckling analysis of cantilever pultruded I-sections using 𝐴𝑁𝑆𝑌𝑆 ®
buckling analysis of cantilever pultruded I-sections using 𝐴𝑁𝑆𝑌𝑆 ®
 
Analysis of elliptical steel tubular section with frp
Analysis of elliptical steel tubular section with frpAnalysis of elliptical steel tubular section with frp
Analysis of elliptical steel tubular section with frp
 
Cq31401405
Cq31401405Cq31401405
Cq31401405
 
Flexural behavioural study on rc beam with
Flexural behavioural study on rc beam withFlexural behavioural study on rc beam with
Flexural behavioural study on rc beam with
 
yarn manu. carding,blowroom
yarn manu. carding,blowroomyarn manu. carding,blowroom
yarn manu. carding,blowroom
 
Effect of Diaphragms in Axially Loaded Timber Hollow.pptx
Effect of Diaphragms in Axially Loaded Timber Hollow.pptxEffect of Diaphragms in Axially Loaded Timber Hollow.pptx
Effect of Diaphragms in Axially Loaded Timber Hollow.pptx
 
Michael Berkson Summer REU Extended Abstract
Michael Berkson Summer REU Extended AbstractMichael Berkson Summer REU Extended Abstract
Michael Berkson Summer REU Extended Abstract
 
Young’s modulus 2 and resistivity
Young’s modulus 2 and resistivityYoung’s modulus 2 and resistivity
Young’s modulus 2 and resistivity
 
Composite materials lecture
Composite materials lectureComposite materials lecture
Composite materials lecture
 

Mesh Quality

  • 1. Mesh Quality Julien Dompierre julien@cerca.umontreal.ca ´ Centre de Recherche en Calcul Applique (CERCA) ´ ´ Ecole Polytechnique de Montreal Mesh Quality – p. 1/331
  • 2. Authors • Research professionals • Julien Dompierre • Paul Labbé • Marie-Gabrielle Vallet • Professors • François Guibault • Jean-Yves Trépanier • Ricardo Camarero Mesh Quality – p. 2/331
  • 3. References – 1 J. D OMPIERRE , P. L ABBÉ , M.-G. VALLET, F. G UIBAULT AND R. C AMARERO , Critères de qualité pour les maillages simpliciaux. in Maillage et adaptation, Hermès, October 2001, Paris, pages 311–348. Mesh Quality – p. 3/331
  • 4. References – 2 A. L IU and B. J OE, Relationship between Tetrahedron Shape Measures, Bit, Vol. 34, pages 268–287, (1994). Mesh Quality – p. 4/331
  • 5. References – 3 P. L ABBÉ, J. D OMPIERRE, M.-G. VALLET, F. G UIBAULT and J.-Y. T RÉPANIER, A Universal Measure of the Conformity of a Mesh with Respect to an Anisotropic Metric Field, Submitted to Int. J. for Numer. Meth. in Engng, (2003). Mesh Quality – p. 5/331
  • 6. References – 4 P. L ABBÉ, J. D OMPIERRE, M.-G. VALLET, F. G UIBAULT and J.-Y. T RÉPANIER, A Measure of the Conformity of a Mesh to an Anisotropic Metric, Tenth International Meshing Roundtable, Newport Beach, CA, pages 319–326, (2001). Mesh Quality – p. 6/331
  • 7. References – 5 P.-L. G EORGE AND H. B O - ROUCHAKI , Triangulation de Delaunay et maillage, appli- cations aux éléments finis. Hermès, 1997, Paris. This book is available in En- glish. Mesh Quality – p. 7/331
  • 8. References – 6 P. J. F REY AND P.-L. G EORGE, Maillages. Ap- plications aux éléments finis. Hermès, 1999, Paris. This book is available in English. Mesh Quality – p. 8/331
  • 9. Table of Contents 1. Introduction 8. Non-Simplicial 2. Simplex Definition Elements 3. Degeneracies of 9. Shape Quality Simplices Visualization 4. Shape Quality of 10. Shape Quality Simplices Equivalence 5. Formulae for Sim- 11. Mesh Quality and plices Optimization 6. Voronoi, Delaunay 12. Size Quality of and Riemann Simplices 7. Shape Quality and 13. Universal Quality Delaunay 14. Conclusions Mesh Quality – p. 9/331
  • 10. Introduction and Justifications We work on mesh generation, mesh adaptation and mesh optimization. How can we choose the configuration that produces the best triangles ? A triangle shape quality measure is needed. Mesh Quality – p. 10/331
  • 11. Face Flipping How can we choose the configuration that produces the best tetrahedra ? A tetrahedron shape quality measure is needed. Mesh Quality – p. 11/331
  • 12. Edge Swapping S4 S3 S4 S3 S5 S5 A A B B S2 S2 S1 S1 How can we choose the configuration that produces the best tetrahedra ? A tetrahedron shape quality measure is needed. Mesh Quality – p. 12/331
  • 13. Mesh Optimization • Let O1 and O2 , two three-dimensional unstructured tetrahedral mesh Optimizers. Mesh Quality – p. 13/331
  • 14. Mesh Optimization • Let O1 and O2 , two three-dimensional unstructured tetrahedral mesh Optimizers. • What is the norm O of a mesh optimizer ? Mesh Quality – p. 13/331
  • 15. Mesh Optimization • Let O1 and O2 , two three-dimensional unstructured tetrahedral mesh Optimizers. • What is the norm O of a mesh optimizer ? • How can it be asserted that O1 > O2 ? Mesh Quality – p. 13/331
  • 16. It’s Obvious ! • Let B be a benchmark. Mesh Quality – p. 14/331
  • 17. It’s Obvious ! • Let B be a benchmark. • Let M1 = O1 (B) be the optimized mesh obtained with the mesh optimizer O1 . Mesh Quality – p. 14/331
  • 18. It’s Obvious ! • Let B be a benchmark. • Let M1 = O1 (B) be the optimized mesh obtained with the mesh optimizer O1 . • Let M2 = O2 (B) be the optimized mesh obtained with the mesh optimizer O2 . Mesh Quality – p. 14/331
  • 19. It’s Obvious ! • Let B be a benchmark. • Let M1 = O1 (B) be the optimized mesh obtained with the mesh optimizer O1 . • Let M2 = O2 (B) be the optimized mesh obtained with the mesh optimizer O2 . • Common sense says : “The proof is in the pudding”. Mesh Quality – p. 14/331
  • 20. It’s Obvious ! • Let B be a benchmark. • Let M1 = O1 (B) be the optimized mesh obtained with the mesh optimizer O1 . • Let M2 = O2 (B) be the optimized mesh obtained with the mesh optimizer O2 . • Common sense says : “The proof is in the pudding”. • If M1 > M2 then O1 > O2 . Mesh Quality – p. 14/331
  • 21. Benchmarks for Mesh Optimization J. D OMPIERRE, P. L ABBÉ, F. G UIBAULT and R. C AMARERO. Proposal of Benchmarks for 3D Unstructured Tetrahedral Mesh Optimization. 7th International Meshing Roundtable, Dearborn, MI, October 1998, pages 459–478. Mesh Quality – p. 15/331
  • 22. The Trick... • Because the norm O of a mesh optimizer is unknown, the comparison of two optimizers is replaced by the comparison of two meshes. Mesh Quality – p. 16/331
  • 23. The Trick... • Because the norm O of a mesh optimizer is unknown, the comparison of two optimizers is replaced by the comparison of two meshes. • What is the norm M of a mesh ? Mesh Quality – p. 16/331
  • 24. The Trick... • Because the norm O of a mesh optimizer is unknown, the comparison of two optimizers is replaced by the comparison of two meshes. • What is the norm M of a mesh ? • How can we assert that M1 > M2 ? Mesh Quality – p. 16/331
  • 25. The Trick... • Because the norm O of a mesh optimizer is unknown, the comparison of two optimizers is replaced by the comparison of two meshes. • What is the norm M of a mesh ? • How can we assert that M1 > M2 ? • This is what you will know soon, or you money back ! Mesh Quality – p. 16/331
  • 26. What to Retain • This lecture is about the quality of the elements of a mesh and the quality of a whole mesh. Mesh Quality – p. 17/331
  • 27. What to Retain • This lecture is about the quality of the elements of a mesh and the quality of a whole mesh. • The concept of element quality is necessary for the algorithms of egde and face swapping. Mesh Quality – p. 17/331
  • 28. What to Retain • This lecture is about the quality of the elements of a mesh and the quality of a whole mesh. • The concept of element quality is necessary for the algorithms of egde and face swapping. • The concept of mesh quality is necessary to do research on mesh optimization. Mesh Quality – p. 17/331
  • 29. Table of Contents 1. Introduction 8. Non-Simplicial 2. Simplex Definition Elements 3. Degeneracies of 9. Shape Quality Simplices Visualization 4. Shape Quality of 10. Shape Quality Simplices Equivalence 5. Formulae for Simplices 11. Mesh Quality and 6. Voronoi, Delaunay and Optimization Riemann 12. Size Quality of 7. Shape Quality and Simplices Delaunay 13. Universal Quality 14. Conclusions Mesh Quality – p. 18/331
  • 30. Definition of a Simplex Meshes in two and three dimensions are made of polygons or polyhedra named elements. Mesh Quality – p. 19/331
  • 31. Definition of a Simplex Meshes in two and three dimensions are made of polygons or polyhedra named elements. The most simple amongst them, the simplices, are those which have the minimal number of vertices. Mesh Quality – p. 19/331
  • 32. Definition of a Simplex Meshes in two and three dimensions are made of polygons or polyhedra named elements. The most simple amongst them, the simplices, are those which have the minimal number of vertices. The segment in one dimension. Mesh Quality – p. 19/331
  • 33. Definition of a Simplex Meshes in two and three dimensions are made of polygons or polyhedra named elements. The most simple amongst them, the simplices, are those which have the minimal number of vertices. The segment in one dimension. The triangle in two dimensions. Mesh Quality – p. 19/331
  • 34. Definition of a Simplex Meshes in two and three dimensions are made of polygons or polyhedra named elements. The most simple amongst them, the simplices, are those which have the minimal number of vertices. The segment in one dimension. The triangle in two dimensions. The tetrahedron in three dimensions. Mesh Quality – p. 19/331
  • 35. Definition of a Simplex Meshes in two and three dimensions are made of polygons or polyhedra named elements. The most simple amongst them, the simplices, are those which have the minimal number of vertices. The segment in one dimension. The triangle in two dimensions. The tetrahedron in three dimensions. The hypertetrahedron in four dimensions. Mesh Quality – p. 19/331
  • 36. Definition of a Simplex Meshes in two and three dimensions are made of polygons or polyhedra named elements. The most simple amongst them, the simplices, are those which have the minimal number of vertices. The segment in one dimension. The triangle in two dimensions. The tetrahedron in three dimensions. The hypertetrahedron in four dimensions. Quadrilaterals, pyramids, prisms, hexahedra and other such aliens are named non-simplicial elements. Mesh Quality – p. 19/331
  • 37. Definition of a d-Simplex in Rd Let d + 1 points Pj = (p1j , p2j , . . . , pdj ) ∈ Rd , 1 ≤ j ≤ d + 1, not in the same hyperplane, id est, such that the matrix of order d + 1,   p11 p12 · · · p1,d+1  p21 p22 · · · p2,d+1     . .  A= . . . .. . . . .  .    pd1 pd2 · · · pd,d+1  1 1 ··· 1 be invertible. The convex hull of the points Pj is named the d-simplex of points Pj . Mesh Quality – p. 20/331
  • 38. A Simplex Generates Rd Any point X ∈ Rd , with Cartesian coordinates (xi )d , is i=1 characterized by the d + 1 scalars λj = λj (X) defined as solution of the linear system  d+1     pij λj = xi for 1 ≤ i ≤ d,  j=1 d+1     λj = 1,  j=1 whose matrix is A. Mesh Quality – p. 21/331
  • 39. What to Retain In two dimensions, the simplex is a triangle. Mesh Quality – p. 22/331
  • 40. What to Retain In two dimensions, the simplex is a triangle. In three dimensions, the simplex is a tetrahedron. Mesh Quality – p. 22/331
  • 41. What to Retain In two dimensions, the simplex is a triangle. In three dimensions, the simplex is a tetrahedron. The d + 1 vertices of a simplex in Rd give d vectors that form a base of Rd . Mesh Quality – p. 22/331
  • 42. What to Retain In two dimensions, the simplex is a triangle. In three dimensions, the simplex is a tetrahedron. The d + 1 vertices of a simplex in Rd give d vectors that form a base of Rd . The coordinates λj (X) of a point X ∈ Rd in the base generated by the simplex are the barycentric coordinates. Mesh Quality – p. 22/331
  • 43. Table of Contents 1. Introduction 8. Non-Simplicial 2. Simplex Definition Elements 3. Degeneracies of 9. Shape Quality Simplices Visualization 4. Shape Quality of 10. Shape Quality Simplices Equivalence 5. Formulae for Simplices 11. Mesh Quality and 6. Voronoi, Delaunay and Optimization Riemann 12. Size Quality of 7. Shape Quality and Simplices Delaunay 13. Universal Quality 14. Conclusions Mesh Quality – p. 23/331
  • 44. Degeneracy of Simplices A d-simplex made of d + 1 vertices Pj is degenerate if its vertices are located in the same hyperplane, id est, if the matrix A is not invertible. Mesh Quality – p. 24/331
  • 45. Degeneracy of Simplices A d-simplex is degenerate if its d + 1 vertices do not generate the space Rd . Mesh Quality – p. 25/331
  • 46. Degeneracy of Simplices A d-simplex is degenerate if its d + 1 vertices do not generate the space Rd . Such is the case if the d + 1 vertices are located in a space of dimension lower than d. Mesh Quality – p. 25/331
  • 47. Degeneracy of Simplices A d-simplex is degenerate if its d + 1 vertices do not generate the space Rd . Such is the case if the d + 1 vertices are located in a space of dimension lower than d. A triangle is degenerate if its vertices are collinear or collapsed. Mesh Quality – p. 25/331
  • 48. Degeneracy of Simplices A d-simplex is degenerate if its d + 1 vertices do not generate the space Rd . Such is the case if the d + 1 vertices are located in a space of dimension lower than d. A triangle is degenerate if its vertices are collinear or collapsed. A tetrahedron is degenerate if its vertices are coplanar, collinear or collapsed. Mesh Quality – p. 25/331
  • 49. Degeneracy of Simplices A d-simplex is degenerate if its d + 1 vertices do not generate the space Rd . Such is the case if the d + 1 vertices are located in a space of dimension lower than d. A triangle is degenerate if its vertices are collinear or collapsed. A tetrahedron is degenerate if its vertices are coplanar, collinear or collapsed. Nota bene : Strictly speaking, accordingly to the definition, a degenerate simplex is no longer a simplex. Mesh Quality – p. 25/331
  • 50. Degeneracy Criterion A d-simplex is degenerate if its matrix A is not invertible. A matrix is not invertible if its determinant is null. Mesh Quality – p. 26/331
  • 51. Degeneracy Criterion A d-simplex is degenerate if its matrix A is not invertible. A matrix is not invertible if its determinant is null. The size of a simplex is its area in two dimensions and its volume in three dimensions. Mesh Quality – p. 26/331
  • 52. Degeneracy Criterion A d-simplex is degenerate if its matrix A is not invertible. A matrix is not invertible if its determinant is null. The size of a simplex is its area in two dimensions and its volume in three dimensions. The size of a d-simplex K made of d + 1 vertices Pj is given by size(K) = det(A)/d!. Mesh Quality – p. 26/331
  • 53. Degeneracy Criterion A d-simplex is degenerate if its matrix A is not invertible. A matrix is not invertible if its determinant is null. The size of a simplex is its area in two dimensions and its volume in three dimensions. The size of a d-simplex K made of d + 1 vertices Pj is given by size(K) = det(A)/d!. A triangle is degenerate if its area is null. Mesh Quality – p. 26/331
  • 54. Degeneracy Criterion A d-simplex is degenerate if its matrix A is not invertible. A matrix is not invertible if its determinant is null. The size of a simplex is its area in two dimensions and its volume in three dimensions. The size of a d-simplex K made of d + 1 vertices Pj is given by size(K) = det(A)/d!. A triangle is degenerate if its area is null. A tetrahedron is degenerate if its volume is null. Mesh Quality – p. 26/331
  • 55. Taxonomy of Degenerate Simplices This taxonomy is based on the different possible degenerate states of the simplices. Mesh Quality – p. 27/331
  • 56. Taxonomy of Degenerate Simplices This taxonomy is based on the different possible degenerate states of the simplices. There are three cases of degenerate triangles. Mesh Quality – p. 27/331
  • 57. Taxonomy of Degenerate Simplices This taxonomy is based on the different possible degenerate states of the simplices. There are three cases of degenerate triangles. There are ten cases of degenerate tetrahedra. Mesh Quality – p. 27/331
  • 58. Taxonomy of Degenerate Simplices This taxonomy is based on the different possible degenerate states of the simplices. There are three cases of degenerate triangles. There are ten cases of degenerate tetrahedra. In this classification, the four symbols , , and stand for vertices of multiplicity simple, double, triple and quadruple respectively. Mesh Quality – p. 27/331
  • 59. 1 – The Cap Name h −→ 0 h=0 C h Cap A B A C B Degenerate edges : None Radius of the smallest circumcircle : ∞ Mesh Quality – p. 28/331
  • 60. 2 – The Needle Name h −→ 0 h=0 C h Needle A B A,C B Degenerate edges : AC Radius of the smallest circumcircle : hmax /2 Mesh Quality – p. 29/331
  • 61. 3 – The Big Crunch Name h −→ 0 h=0 C h h B Big A h A,B,C Crunch Degenerate edges : All Radius of the smallest circumcircle : 0 The Big Crunch is the theory opposite of the Big Bang. Mesh Quality – p. 30/331
  • 62. Degeneracy of Tetrahedra There is one case of degeneracy resulting in four collapsed vertices. There are five cases of degeneracy resulting in four collinear vertices. There are four cases of degeneracy resulting in four coplanar vertices. D D d A C A a C b B B c Mesh Quality – p. 31/331
  • 63. 1 – The Fin Name h −→ 0 h=0 D h D A C A C Fin B B Degenerate edges : None Degenerate faces : One cap Radius of the smallest circumsphere : ∞ Mesh Quality – p. 32/331
  • 64. 2 – The Cap Name h −→ 0 h=0 D Cap A h C A D C B B Degenerate edges : None Degenerate faces : None Radius of the smallest circumsphere : ∞ Mesh Quality – p. 33/331
  • 65. 3 – The Sliver Name h −→ 0 h=0 D h C Sliver A C A D B B Degenerate edges : None Degenerate faces : None Radius of the smallest circumsphere : rABC or ∞ Mesh Quality – p. 34/331
  • 66. 4 – The Wedge Name h −→ 0 h=0 D h C, D Wedge A C A B B Degenerate edges : CD Degenerate faces : Two needles Radius of the smallest circumsphere : rABC Mesh Quality – p. 35/331
  • 67. 5 – The Crystal Name h −→ 0 h=0 D A h Crystal h C A B D C B Degenerate edges : None Degenerate faces : Four caps Radius of the smallest circumsphere : ∞ Mesh Quality – p. 36/331
  • 68. 6 – The Spindle Name h −→ 0 h=0 D A h A B, D C Spindle h C B Degenerate edges : BD Degenerate faces : Two caps and two needles Radius of the smallest circumsphere : ∞ Mesh Quality – p. 37/331
  • 69. 7 – The Splitter Name h −→ 0 h=0 D h C Splitter A A D B, C h B Degenerate edges : BC Degenerate faces : Two caps and two needles Radius of the smallest circumsphere : ∞ Mesh Quality – p. 38/331
  • 70. 8 – The Slat Name h −→ 0 h=0 D h C Slat h A, D B, C A B Degenerate edges : AD and BC Degenerate faces : Four needles Radius of the smallest circumsphere : hmax /2 Mesh Quality – p. 39/331
  • 71. 9 – The Needle Name h −→ 0 h=0 D h h hC A B, C, D Needle A B Degenerate edges : BC, CD and DB Degenerate faces : Three needles and one Big Crunch Radius of the smallest circumsphere : hmax /2 Mesh Quality – p. 40/331
  • 72. 10 – The Big Crunch Name h −→ 0 h=0 D Big A hh C h h A, B, C, D Crunch h Bh Degenerate edges : All Degenerate faces : Four Big Crunches Radius of the smallest circumsphere : 0 Mesh Quality – p. 41/331
  • 73. What to Retain A triangle is degenerate if its vertices are collinear or collapsed, hence if its area is null. Mesh Quality – p. 42/331
  • 74. What to Retain A triangle is degenerate if its vertices are collinear or collapsed, hence if its area is null. There are three cases of degeneracy of triangles. Mesh Quality – p. 42/331
  • 75. What to Retain A triangle is degenerate if its vertices are collinear or collapsed, hence if its area is null. There are three cases of degeneracy of triangles. A tetrahedron is degenerate if its vertices are coplanar, collinear or collapsed, hence if its volume is null. Mesh Quality – p. 42/331
  • 76. What to Retain A triangle is degenerate if its vertices are collinear or collapsed, hence if its area is null. There are three cases of degeneracy of triangles. A tetrahedron is degenerate if its vertices are coplanar, collinear or collapsed, hence if its volume is null. There are ten cases of degeneracy of tetrahedra. Mesh Quality – p. 42/331
  • 77. Table of Contents 1. Introduction 8. Non-Simplicial 2. Simplex Definition Elements 3. Degeneracies of 9. Shape Quality Simplices Visualization 4. Shape Quality of 10. Shape Quality Simplices Equivalence 5. Formulae for Simplices 11. Mesh Quality and 6. Voronoi, Delaunay and Optimization Riemann 12. Size Quality of 7. Shape Quality and Simplices Delaunay 13. Universal Quality 14. Conclusions Mesh Quality – p. 43/331
  • 78. Shape Quality of Simplices An usual method used to quantify the quality of a mesh is through the quality of the elements of that mesh. Mesh Quality – p. 44/331
  • 79. Shape Quality of Simplices An usual method used to quantify the quality of a mesh is through the quality of the elements of that mesh. A criterion usually used to quantify the quality of an element is the shape measure. Mesh Quality – p. 44/331
  • 80. Shape Quality of Simplices An usual method used to quantify the quality of a mesh is through the quality of the elements of that mesh. A criterion usually used to quantify the quality of an element is the shape measure. This section is a guided tour of the shape measures used for simplices. Mesh Quality – p. 44/331
  • 81. The Regular Simplex Definition : An element is regular if it maximizes its measure for a given measure of its boundary. Mesh Quality – p. 45/331
  • 82. The Regular Simplex Definition : An element is regular if it maximizes its measure for a given measure of its boundary. The equilateral triangle is regular because it maximizes its area for a given perimeter. Mesh Quality – p. 45/331
  • 83. The Regular Simplex Definition : An element is regular if it maximizes its measure for a given measure of its boundary. The equilateral triangle is regular because it maximizes its area for a given perimeter. The equilateral tetrahedron is regular because it maximizes its volume for a given surface of its faces. Mesh Quality – p. 45/331
  • 84. Simplicial Shape Measure Definition A : A simplicial shape measure is a continuous function that evaluates the shape of a simplex. It must be invariant under translation, rotation, reflection and uniform scaling of the simplex. A shape measure is called valid if it is maximal only for the regular simplex and if it is minimal for all degenerate simplices. Simplicial shape measures are scaled to the interval [0, 1], and are 1 for the regular simplex and 0 for a degenerate simplex. Mesh Quality – p. 46/331
  • 85. Remarks The invariance under translation, rotation and reflection means that the simplicial shape measures must be independent of the coordinates system. Mesh Quality – p. 47/331
  • 86. Remarks The invariance under translation, rotation and reflection means that the simplicial shape measures must be independent of the coordinates system. The invariance under a valid uniform scaling means that the simplicial shape measures must be dimensionless (independent of the unit system). Mesh Quality – p. 47/331
  • 87. Remarks The invariance under translation, rotation and reflection means that the simplicial shape measures must be independent of the coordinates system. The invariance under a valid uniform scaling means that the simplicial shape measures must be dimensionless (independent of the unit system). The continuity means that the simplicial shape measures must change continuously in function of the coordinates of the vertices of the simplex. Mesh Quality – p. 47/331
  • 88. The Radius Ratio The radius ratio of a simplex K is a shape measure defined as ρ = d ρK /rK , where ρK and rK are the radius of the incircle and circumcircle of K (insphere and circumsphere in 3D), and where d is the dimension of space. K ρK rK Mesh Quality – p. 48/331
  • 89. The Mean Ratio Let R(r1 , r2 , r3 [, r4 ]) be an equilateral simplex having the same [area|volume] than the simplex K(P1 , P2 , P3 [, P4 ]). Let N be the matrix of transformation from R to K, i.e. Pi = N ri + b, 1 ≤ i ≤ [3|4], where b is a translation vector. s y K K = NR + b R r b x Mesh Quality – p. 49/331
  • 90. The Mean Ratio Then, the mean ratio η of the simplex K is the ratio of the geometric mean over the algebraic means of the eigenvalues λ1 , λ2 [,λ3 ] of the matrix N T N .  √ √  2 λ1 λ2  2 4 3 SK d   λ +λ = in 2D, d λi   1 2 2 1≤i<j≤3 Lij i=1  η= = d  1 λi  3 √λ 1 λ 2 λ 3   3 12 3 9VK2 d   i=1  λ +λ +λ = L 2 in 3D. 1 2 3 1≤i<j≤4 ij Mesh Quality – p. 50/331
  • 91. The Condition Number F ORMAGGIA and P EROTTO (2000) use the inverse of the condition number of the matrix. min λiλ1 i κ= = , max λi λd i if the eigenvalues are sorted in increasing order. Mesh Quality – p. 51/331
  • 92. The Frobenius Norm Freitag and Knupp (1999) use the Frobenius norm of the matrix N = AW −1 to define a shape measure. d d κ= = , tr(N T N )tr((N T N )−1 ) d d λi λ−1 i i=1 i=1 where the λi are the eigenvalues of the tensor N T N . Mesh Quality – p. 52/331
  • 93. The Minimum of Solid Angles The simplicial shape measure θmin based on the minimum of solid angles of the d-simplex is defined by θmin = α−1 min θi , 1≤i≤d+1 The coefficient α is the value of each solid angle of the regular d-simplex, given by α = π/3 in two dimensions √ and α = 6 arcsin 3/3 − π in three dimensions. Mesh Quality – p. 53/331
  • 94. The sin of θmin From a numerical point of view, a less expensive simplicial shape measure is the sin of the minimum solid angle. This avoids the computation of the arcsin(·) function in the computation of θi in 2D and θi in 3D. σmin = β −1 min σi , 1≤i≤d+1 where σi = sin(θi ) in 2D and σi = sin(θi /2) in 3D. β is the value of σi for all solid angles of the regular simplex, given √ √ by β = sin(α) = 3/2 in 2D and β = sin(α/2) = 6/9 in 3D. Mesh Quality – p. 54/331
  • 95. Face Angles We can define a shape measure based on the minimum of the twelve angles of the four faces of a tetrahedron. This angle is π/3 for the regular tetrahedron. But this shape measure is not valid according to Definition A because it is insensitive to degenerate tetrahedra that do not have degenerate faces (the sliver and the cap). Mesh Quality – p. 55/331
  • 96. Dihedral Angles The dihedral angle is the angle between the intersection of two adjacent faces to an edge with the perpendicular plane of the edge. Pj ϕij Pi The minimum of the six dihedral angles ϕmin is used as a shape measure. Mesh Quality – p. 56/331
  • 97. Dihedral Angles αϕmin = min ϕij = min (π − arccos (nij1 · nij2 )) , 1≤i<j≤4 1≤i<j≤4 where nij1 and nij2 are the normal to the adjacent faces of the edge Pi Pj , and where α = π − arccos(−1/3) is the value of the six dihedral angles of the regular tetrahedron. But this shape measure is not valid according to Definition A. The smallest dihedral angles of the needle, the spindle and the crystal can be as large as π/3. Mesh Quality – p. 57/331
  • 98. The Interpolation Error Coefficient In finite element, the interpolation error of a function over an element is bounded by a coefficient times the semi-norm of the function. This coefficient is the ratio DK /̺K where DK is the diameter of the element K and ̺K is the roundness of the element K.  √ ρK  2 3  in 2 D, hmax γ=  2√6 ρK in 3 D.  hmax Mesh Quality – p. 58/331
  • 99. The Edge Ratio Ratio of the smallest edge over the tallest. r = hmin /hmax . The edge ratio r is not a valid shape measure according to Definition A because it does not vanish for some degenerate simplices. In 2D, it can be as large as 1/2 for √ the cap. In 3D, it can be as large as 2/2 for the sliver, 1/2 √ for the fin, 3/3 for the cap and 1/3 for the crystal. Mesh Quality – p. 59/331
  • 100. Other Shape Measure – 1 hmax /rK , the ratio of the diameter of the tetrahedron over the circumradius, in B AKER, (1989). This is not a valid shape measure. hmin /rK , the ratio of the smallest edge of the tetrahedron over the circumradius, in M ILLER et al (1996). This is not a valid shape measure. VK /rK 3 , the ratio of the volume of the tetrahedron over the circumradius, in M ARCUM et W EATHERILL, (1995). Mesh Quality – p. 60/331
  • 101. Other Shape Measure – 2 4 4 2 −3 VK i=1 Si , the ratio of the volume of the tetrahedron over the area of its faces, in D E C OUGNY et al (1990). The evaluation of this shape measure, and its validity, are a complex problem for tetrahedra that degenerate in four collinear vertices. −3 VK 1≤i<j≤4 Lij , the ratio of the volume of the tetrahedron over the average of its edges, in DANNELONGUE and TANGUY (1991), Z AVATTIERI et al (1996) and W EATHERILL et al (1993). Mesh Quality – p. 61/331
  • 102. Other Shape Measure – 3 2 VK Lij − L12 L34 − L13 L24 1≤i<j≤4 −3/2 −L14 L23 + L2 ij 1≤i<j≤4 the ratio of the volume of the tetrahedron over a sum, at the power three halfs, of many terms homogeneous to the square of edge lenghts, in B ERZINS (1998). Mesh Quality – p. 62/331
  • 103. Other Shape Measure – 4 −3 VK L2 1≤i<j≤4 ij , the ratio of the volume of the tetrahedron over the quadratic average of the six edges, in G RAICHEN et al (1991). And so on... This list is surely not exhaustive. Mesh Quality – p. 63/331
  • 104. There Exists an Infinity of Shape Measures If µ and ν are two valid shape measures, if c, d ∈ R+ , then µc , c(µ−1)/µ with c > 1, αµc + (1 − α)ν d with α ∈ [0, 1], µc ν d are also valid simplicial shape measures. Mesh Quality – p. 64/331
  • 105. What to Retain The regular simplex is the equilateral one, ie, where all its edges have the same length. Mesh Quality – p. 65/331
  • 106. What to Retain The regular simplex is the equilateral one, ie, where all its edges have the same length. A shape measures evaluates the ratio to equilaterality. Mesh Quality – p. 65/331
  • 107. What to Retain The regular simplex is the equilateral one, ie, where all its edges have the same length. A shape measures evaluates the ratio to equilaterality. A non valid shape measure does not vanish for all degenerate simplices. Mesh Quality – p. 65/331
  • 108. What to Retain The regular simplex is the equilateral one, ie, where all its edges have the same length. A shape measures evaluates the ratio to equilaterality. A non valid shape measure does not vanish for all degenerate simplices. There exists an infinity of valid shape measures. Mesh Quality – p. 65/331
  • 109. What to Retain The regular simplex is the equilateral one, ie, where all its edges have the same length. A shape measures evaluates the ratio to equilaterality. A non valid shape measure does not vanish for all degenerate simplices. There exists an infinity of valid shape measures. The goal of research is not to find an other one way better than the other ones. Mesh Quality – p. 65/331
  • 110. Table of Contents 1. Introduction 8. Non-Simplicial 2. Simplex Definition Elements 3. Degeneracies of 9. Shape Quality Simplices Visualization 4. Shape Quality of 10. Shape Quality Simplices Equivalence 5. Formulae for Simplices 11. Mesh Quality and 6. Voronoi, Delaunay and Optimization Riemann 12. Size Quality of 7. Shape Quality and Simplices Delaunay 13. Universal Quality 14. Conclusions Mesh Quality – p. 66/331
  • 111. Formulae for the Triangle A triangle is completely defined by the knowledge of the length of its three edges. Quantities such that inradius, circumradius, angles, area, etc, can be written in function of the edge lengths of the triangle. Let K be a non degenerate triangle of vertices P1 , P2 and P3 . The lengths of the edges Pi Pj of K are denoted Lij = Pj − Pi , 1 ≤ i < j ≤ 3. Mesh Quality – p. 67/331
  • 112. The Half-Perimeter The half-perimeter pK is given by (L12 + L13 + L23 ) pK = . 2 Mesh Quality – p. 68/331
  • 113. Heron’s Formula The area SK of a triangle can also be written in function of the edge lengths with Heron’s formula : 2 SK = pK (pK − L12 )(pK − L13 )(pK − L23 ). Mesh Quality – p. 69/331
  • 114. Radius of the Incircle The radius ρK of the incircle of the triangle K is given by SK ρK = . pK Mesh Quality – p. 70/331
  • 115. Radius of the Circumscribed Circle The radius rK of the circumcircle of the triangle K is given by L12 L13 L23 rK = . 4SK Mesh Quality – p. 71/331
  • 116. Element Diameter The diameter of an element is the biggest Euclidean distance between two points of an element. For a triangle, this is also the length of the biggest edge hmax hmax = max(L12 , L13 , L23 ), The length of the smallest edge is denoted hmin hmin = min(L12 , L13 , L23 ). Mesh Quality – p. 72/331
  • 117. Solid Angle The angle θi at vertex Pi of triangle K is the arc length obtained by projecting the edge of the triangle opposite to Pi on a unitary circle centerered at Pi . The angle can be written in function of the edge lengths as −1 θi = arcsin 2SK Lij Lik . j,k=i 1≤j<k≤3 Mesh Quality – p. 73/331
  • 118. Formulae for the Tetrahedron A tetrahedron is completely defined by the knowledge of the length of its six edges. Quantities such that inradius, circumradius, angles, volume, etc, can be written in function of the edge lengths of the tetrahedron. Mesh Quality – p. 74/331
  • 119. Formulae for the Tetrahedron Let K be a non degenerate tetrahedron of vertices P1 , P2 , P3 and P4 . The lengths of the edges Pi Pj of K are denoted Lij = Pj − Pi , 1 ≤ i < j ≤ 4. The area of the four faces of the tetrahedron, △P2 P3 P4 , △P1 P3 P4 , △P1 P2 P4 and △P1 P2 P3 , are denoted by S1 , S2 , S3 and S4 . Finally, VK is the volume of the tetrahedron K. Mesh Quality – p. 75/331
  • 120. 3D “Heron’s” Formula Let a, b, c, e, f and g be the length of the six edges of the tetrahedron such that the edges a, b and c are connected to the same vertex, and such that e is the opposite edge of a, f is opposite of b and g is the opposite of c. The volume VK is then 2 144VK = 4a2 b2 c2 + (b2 + c2 − e2 ) (c2 + a2 − f 2 ) (a2 + b2 − g 2 ) 2 2 − a2 (b2 + c2 − e2 ) − b2 (c2 + a2 − f 2 ) 2 − c2 (a2 + b2 − g 2 ) . Mesh Quality – p. 76/331
  • 121. Radius of the Insphere The radius ρK of the insphere of the tetrahedron K is given by 3VK ρK = . S1 + S2 + S3 + S4 Mesh Quality – p. 77/331
  • 122. Radius of the Circumsphere The radius rK of the circumsphere of the tetrahedron K is given by (a + b + c)(a + b − c)(a + c − b)(b + c − a) rK = . 24VK where a = L12 L34 , b = L13 L24 and c = L14 L23 are the product of the length of the opposite edges of K (two edges are opposite if they do not share a vertex. Mesh Quality – p. 78/331
  • 123. Element Diameter The diameter of an element is the biggest Euclidean distance between two points of an element. For a tetrahedron, this is also the length of the biggest edge hmax hmax = max(L12 , L13 , L14 , L23 , L24 , L34 ), The length of the smallest edge is denoted hmin hmin = min(L12 , L13 , L14 , L23 , L24 , L34 ). Mesh Quality – p. 79/331
  • 124. Solid Angle The solid angle θi at vertex Pi of the tetrahedron K, is the area of the spherical sector obtained by projecting the face of the tetrahedron opposite to Pi on a unitary sphere centerered at Pi . P4 P1 θ1 P3 P2 Mesh Quality – p. 80/331
  • 125. Solid angle L IU and J OE (1994) gave a formula to compute the solid angle in function of edge lengths : −1/2 θi = 2 arcsin 12VK (Lij + Lik )2 − L2 jk . j,k=i 1≤j<k≤4 Mesh Quality – p. 81/331
  • 126. Table of Contents 1. Introduction 8. Non-Simplicial 2. Simplex Definition Elements 3. Degeneracies of 9. Shape Quality Simplices Visualization 4. Shape Quality of 10. Shape Quality Simplices Equivalence 5. Formulae for Sim- 11. Mesh Quality and plices Optimization 6. Voronoi, Delaunay 12. Size Quality of and Riemann Simplices 7. Shape Quality and 13. Universal Quality Delaunay 14. Conclusions Mesh Quality – p. 82/331
  • 127. Which Is the Most Beautiful Triangle ? Mesh Quality – p. 83/331
  • 128. Which Is the Most Beautiful Triangle ? A Mesh Quality – p. 83/331
  • 129. Which Is the Most Beautiful Triangle ? A B Mesh Quality – p. 83/331
  • 130. If You Chose the Triangle A... Mesh Quality – p. 84/331
  • 131. If You Chose the Triangle A... A You are wrong ! Mesh Quality – p. 84/331
  • 132. If You Chose the Triangle B... Mesh Quality – p. 85/331
  • 133. If You Chose the Triangle B... B You are wrong again ! Mesh Quality – p. 85/331
  • 134. Which Is the Most Beautiful Triangle ? A B None of these answers ! Mesh Quality – p. 86/331
  • 135. Which Is the Most Beautiful Woman ? Mesh Quality – p. 87/331
  • 136. Which Is the Most Beautiful Woman ? A Mesh Quality – p. 87/331
  • 137. Which Is the Most Beautiful Woman ? A B Mesh Quality – p. 87/331
  • 138. You Probably chose... Mesh Quality – p. 88/331
  • 139. You Probably chose... A B Woman A. Mesh Quality – p. 88/331
  • 140. And if One Asked these Gentlemen... Mesh Quality – p. 89/331
  • 141. And if One Asked these Gentlemen... Mesh Quality – p. 89/331
  • 142. These Gentlemen Would Choose... Mesh Quality – p. 90/331
  • 143. These Gentlemen Would Choose... A B Woman B. Mesh Quality – p. 90/331
  • 144. Which Is the Most Beautiful Woman... There is no absolute answer because the question is incomplete. One did not specify who was going to judge the candidates, which was the scale of evaluation, which were the measurements used, etc. Mesh Quality – p. 91/331
  • 145. Which Is the Most Beautiful Triangle ? Mesh Quality – p. 92/331
  • 146. Which Is the Most Beautiful Triangle ? A B Mesh Quality – p. 92/331
  • 147. Which Is the Most Beautiful Triangle ? A B The question is incomplete : It misses a way of measuring the quality of a triangle. Mesh Quality – p. 92/331
  • 148. Voronoi Diagram Georgy Fedoseevich VORO - NOÏ . April 28, 1868, Ukraine – November 20, 1908, War- saw. Nouvelles applications des paramètres continus à la théorie des formes qua- dratiques. Recherches sur les parallélloèdes primitifs. Journal Reine Angew. Math, Vol 134, 1908. Mesh Quality – p. 93/331
  • 149. The Perpendicular Bisector Let S1 and S2 be two vertices in R2 . The perpendicular bisec- d(P, S1 ) P tor M (S1 , S2 ) is the S1 d(P, S2 ) locus of points equi- distant to S1 and S2 . S2 M (S1 , S2 ) = {P ∈ M R2 | d(P, S1 ) = d(P, S2 )}, where d(·, ·) is the Eucli- dean distance between two points of space. Mesh Quality – p. 94/331
  • 150. A Cloud of Vertices Let S = {Si }i=1,...,N be a cloud of N vertices. S2 S11 S9 S10 S5 S6 S4 S8 S1 S7 S12 S3 Mesh Quality – p. 95/331
  • 151. The Voronoi Cell Definition : The Voronoi cell C(Si ) associated to the vertex Si is the locus of points of space which is closer to Si than any other vertex : C(Si ) = {P ∈ R2 | d(P, Si ) ≤ d(P, Sj ), ∀j = i}. C(Si ) Si Mesh Quality – p. 96/331
  • 152. The Voronoi Diagram The set of Voronoi cells associated with all the vertices of the cloud of vertices is called the Voronoi diagram. Mesh Quality – p. 97/331
  • 153. Properties of the Voronoi Diagram The Voronoi cells are polygons in 2D, polyhedra in 3D and N -polytopes in N D. The Voronoi cells are convex. The Voronoi cells cover space without overlapping. Mesh Quality – p. 98/331
  • 154. What to Retain The Voronoi diagrams are partitions of space into cells based on the concept of distance. Mesh Quality – p. 99/331
  • 155. Delaunay Triangulation Boris Nikolaevich D ELONE or D ELAUNAY. 15 mars 1890, Saint Petersbourg — 1980. Sur la sphère vide. À la mé- moire de Georges Voronoi, Bulletin of the Academy of Sciences of the USSR, Vol. 7, pp. 793–800, 1934. Mesh Quality – p. 100/331
  • 156. Triangulation of a cloud of Points The same cloud of points can be triangulated in many different fashions. ... Mesh Quality – p. 101/331
  • 157. Triangulation of a Cloud of Points ... ... Mesh Quality – p. 102/331
  • 158. Triangulation of a Cloud of Points ... ... Mesh Quality – p. 103/331
  • 159. Delaunay Triangulation Among all these fashions, there is one (or maybe many) triangulation of the convex hull of the point cloud that is said to be a Delaunay Triangulation. Mesh Quality – p. 104/331
  • 160. Empty Sphere Criterion of Delaunay Empty sphere criterion : A simplex K satisfies the empty sphere criterion if the open circumscribed ball of the simplex K is empty (ie, does not contain any other vertex of the triangulation). K K Mesh Quality – p. 105/331
  • 161. Violation of the Empty Sphere Criterio A simplex K does not satisfy the empty sphere criterion if the opened circumscribed ball of simplex K is not empty (ie, it contains at least one vertex of the triangulation). K K Mesh Quality – p. 106/331
  • 162. Delaunay Triangulation Delaunay Triangulation : If all the simplices K of a triangulation T satisfy the empty sphere criterion, then the triangulation is said to be a Delaunay triangulation. Mesh Quality – p. 107/331
  • 163. Delaunay Algorithm The circumscri- bed sphere of a simplex has to be S3 computed. S2 This amounts to ρout computing the cen- C ter of a simplex. The center is the point at equal dis- tance to all the vertices of the sim- S1 plex. Mesh Quality – p. 108/331
  • 164. Delaunay Algorithm How can we know if a point P violates the empty sphere criterion for a simplex K ? The center C and the radius ρ of the circumscribed sphere of the simplex K has to be computed. The distance d between the point P and the center C has to be computed. If the distance d is greater than the radius ρ, the point P is not in the circumscribed sphere of the simplex K. Mesh Quality – p. 109/331
  • 165. What to Retain The Voronoi diagram of a cloud of points is a partition of space into cells based on the notion of distance. A Delaunay triangulation of a cloud of points is a triangulation based on the notion of distance. Mesh Quality – p. 110/331
  • 166. Duality Delaunay-Voronoï The Voronoï diagram is the dual of the Delaunay triangulation and vice versa. Mesh Quality – p. 111/331
  • 167. Voronoï and Delaunay in Nature Voronoï diagrams and Delaunay triangulations are not just a mathematician’s whim, they represent structures that can be found in nature. Mesh Quality – p. 112/331
  • 168. Voronoï and Delaunay In Nature Mesh Quality – p. 113/331
  • 169. A Turtle Mesh Quality – p. 114/331
  • 170. A Pineapple Mesh Quality – p. 115/331
  • 171. The Devil’s Tower Mesh Quality – p. 116/331
  • 172. Dry Mud Mesh Quality – p. 117/331
  • 173. Bee Cells Mesh Quality – p. 118/331
  • 174. Dragonfly Wings Mesh Quality – p. 119/331
  • 175. Pop Corn Mesh Quality – p. 120/331
  • 176. Fly Eyes Mesh Quality – p. 121/331
  • 177. Carbon Nanotubes Mesh Quality – p. 122/331
  • 178. Soap Bubbles Mesh Quality – p. 123/331
  • 179. A Geodesic Dome Mesh Quality – p. 124/331
  • 180. Biosphère de Montréal Mesh Quality – p. 125/331
  • 181. Streets of Paris Mesh Quality – p. 126/331
  • 182. Roads in France Mesh Quality – p. 127/331
  • 183. Roads in France Mesh Quality – p. 128/331
  • 184. Where Is this Guy Going ? ! ! A simplicial shape measure is an evaluation of the ratio to equilarity. Mesh Quality – p. 129/331
  • 185. Where Is this Guy Going ? ! ! A simplicial shape measure is an evaluation of the ratio to equilarity. The Voronoï diagram of a cloud of points is a partition of space into cells based on the notion of distance. Mesh Quality – p. 129/331
  • 186. Where Is this Guy Going ? ! ! A simplicial shape measure is an evaluation of the ratio to equilarity. The Voronoï diagram of a cloud of points is a partition of space into cells based on the notion of distance. A Delaunay triangulation of a cloud of points is a triangulation based on the notion of distance. Mesh Quality – p. 129/331
  • 187. Where Is this Guy Going ? ! ! A simplicial shape measure is an evaluation of the ratio to equilarity. The Voronoï diagram of a cloud of points is a partition of space into cells based on the notion of distance. A Delaunay triangulation of a cloud of points is a triangulation based on the notion of distance. The notion of distance can be generalized. Mesh Quality – p. 129/331
  • 188. Where Is this Guy Going ? ! ! A simplicial shape measure is an evaluation of the ratio to equilarity. The Voronoï diagram of a cloud of points is a partition of space into cells based on the notion of distance. A Delaunay triangulation of a cloud of points is a triangulation based on the notion of distance. The notion of distance can be generalized. The notions of shape measure, of Voronoï diagram and of Delaunay triangulation Quality –be Mesh can p. 129/331
  • 189. Nikolai Ivanovich Lobachevsky N IKOLAI I VANOVICH LOBACHEVSKY, 1 décembre 1792, Nizhny Novgorod — 24 février 1856, Kazan. Mesh Quality – p. 130/331
  • 190. János Bolyai J ÁNOS BOLYAI, 15 dé- cembre 1802 à Kolozsvár, Empire Austrichien (Cluj, Roumanie) — 27 janvier 1860 à Marosvásárhely, Empire Austrichien (Tirgu- Mures, Roumanie). Mesh Quality – p. 131/331
  • 191. Bernhard RIEMANN G EORG F RIEDRICH B ERN - HARD RIEMANN, 7 sep- tembre 1826, Hanovre — 20 juillet 1866, Selasca. Über die Hypothesen welche der Geo- metrie zu Grunde liegen. 10 juin 1854. Mesh Quality – p. 132/331
  • 192. Non Euclidean Geometry Riemann has generalized Euclidean geometry in the plane to Riemannian geometry on a surface. He has defined the distance between two points on a surface as the length of the shortest path between these two points (geodesic). He has introduced the Riemannian metric that defines the curvature of space. Mesh Quality – p. 133/331
  • 193. The Metric in the Merriam-Webster Mesh Quality – p. 134/331
  • 194. Definition of a Metric If S is any set, then the function d : S×S → I R is called a metric on S if it satisfies (i) d(x, y) ≥ 0 for all x, y in S ; (ii) d(x, y) = 0 if and only if x = y ; (iii) d(x, y) = d(y, x) for all x, y in S ; (iv) d(x, y) ≤ d(x, z) + d(z, y) for all x, y, z in S. Mesh Quality – p. 135/331
  • 195. The Euclidean Distance is a Metric In the previous definition of a metric, let the set S be I 2 , the function R d : I 2 ×I 2 → I R R R x1 x2 × → (x2 − x1 )2 + (y2 − y1 )2 y1 y2 is a metric on I 2 . R Mesh Quality – p. 136/331
  • 196. Metric Space Mesh Quality – p. 137/331
  • 197. The Scalar Product is a Metric Let a vectorial space with its scalar product ·, · . Then the norm of the scalar product of the difference of two elements of the vectorial space is a metric. d(A, B) = B−A , 1/2 = B − A, B − A , − − 1/2 → → = AB, AB , − T− → → = AB AB. Mesh Quality – p. 138/331
  • 198. The Scalar Product is a Metric If the vectorial space is I 2 , then the norm of the R − → scalar product of the vector AB is the Euclidean distance. 1/2 − T− → → d(A, B) = B − A, B − A = AB AB, T xB − xA xB − xA = , y B − yA y B − yA = (xB − xA )2 + (yB − yA )2 . Mesh Quality – p. 139/331
  • 199. Metric Tensor A metric tensor M is a symmetric positive definite matrix m11 m12 M= in 2D, m12 m22   m11 m12 m13   M =  m12 m22 m23  in 3D. m13 m23 m33 Mesh Quality – p. 140/331
  • 200. Metric Length −→ The length LM (AB) of an edge between vertices A and B in the metric M is given by −→ − − 1/2 → → LM (AB) = AB, AB M , −→ − 1/2 → = AB, M AB , − T − → → = AB M AB. Mesh Quality – p. 141/331
  • 201. Euclidean Length with M = I −→ −→ −→ 1/2 − T − → → LM (AB) = AB, M AB = AB M AB, T xB − xA 1 0 xB − xA = y B − yA 0 1 y B − yA −→ LE (AB) = (xB − xA )2 + (yB − yA )2 . Mesh Quality – p. 142/331
  • 202. αβ Metric Length with M = βγ −→ −→ −→ 1/2 − T − → → LM (AB) = AB, M AB = AB M AB, T xB − xA α β xB − xA = y B − yA β γ y B − yA −→ LE (AB) = α(xB − xA )2 + 2β(xB − xA )(yB − yA ) 2 1/2 +γ(yB − yA ) . Mesh Quality – p. 143/331
  • 203. Length in a Variable Metric In the general sense, the metric tensor M is not constant but varies continuously for every point of space. The length of a parameterized curve γ(t) = {(x(t), y(t), z(t)) , t ∈ [0, 1]} is evaluated in the metric 1 LM (γ) = (γ ′ (t))T M (γ(t)) γ ′ (t) dt, 0 where γ(t) is a point of the curve and γ ′ (t) is the tangent vector of the curve at that point. LM (γ) is always bigger or equal to the geodesic between the end points of the curve. Mesh Quality – p. 144/331
  • 204. Area and Volume in a Metric Area of the triangle K in a metric M : AM (K) = det(M) dA. K Volume of the tetrahedron K in a metric M : VM (K) = det(M) dV. K Mesh Quality – p. 145/331
  • 205. Metric and Delaunay Mesh Mesh Quality – p. 146/331
  • 206. Which is the Best Triangle ? A B The question is incomplete. The way to measure the quality of the triangle is missing. Mesh Quality – p. 147/331
  • 207. Which is the Best Triangle ? A B Mesh Quality – p. 148/331
  • 208. Which is the Best Triangle ? A B Mesh Quality – p. 149/331
  • 209. Example of an Adapted Mesh Adapted mesh and solution for a transonic visquous compressible flow with Mach 0.85 and Reynolds = 5 000. Mesh Quality – p. 150/331
  • 210. Zoom on Boundary Layer–Shock Mesh Quality – p. 151/331
  • 211. What to Retain Beauty, quality and shape are relative notions. Mesh Quality – p. 152/331
  • 212. What to Retain Beauty, quality and shape are relative notions. We first need to define what we want in order to evaluate what we obtained. Mesh Quality – p. 152/331
  • 213. What to Retain Beauty, quality and shape are relative notions. We first need to define what we want in order to evaluate what we obtained. “What we want” is written in the form of metric tensors. Mesh Quality – p. 152/331
  • 214. What to Retain Beauty, quality and shape are relative notions. We first need to define what we want in order to evaluate what we obtained. “What we want” is written in the form of metric tensors. A shape measure is a measure of the equilarity of a simplex in this metric. Mesh Quality – p. 152/331
  • 215. Shape Measure in a Metric First method (constant metric) For a simplex K, evaluate the metric tensor at several points (Gaussian points) and find an averaged metric tensor. Take this averaged metric tensor as constant over the whole simplex and evaluate the shape measure using this metric. Mesh Quality – p. 153/331
  • 216. Shape Measure in a Metric Second method (constant metric) For a simplex K, evaluate the metric tensor at one point (Gaussian point) and take the metric as constant over the whole simplex. Evaluate the shape measure using this metric. Repeat this operation at several points and average the shape measures. This is what is done at INRIA. Mesh Quality – p. 154/331
  • 217. Shape Measure in a Metric Third methode (variable metric) Express the shape measure as a fonction of edge lengths only. Evaluate the length of the edges in the metric and compute the shape measure with these lengths. This is what is done in OORT. Mesh Quality – p. 155/331
  • 218. Shape Measure in a Metric Fourth method (variable metric) Express the shape measure in function of the length of the edges, the area and the volumes. Evaluate the lengths, the area and the volume in the metric. Mesh Quality – p. 156/331
  • 219. Shape Measure in a Metric Fifth method (variable metric) Know how to evaluate quantities such as the radius of the inscribed circle, of the circumscribed circle, the solid angle, etc, in a metric. In the general sense, the triangular inequality is not verified in a variable metric. Neither is the sum of the angles equal to 180 degrees, etc. The evaluation of a shape measure in a variable metric in all its generality is an opened problem. For the moment, it is approximated. Mesh Quality – p. 157/331
  • 220. Table of Contents 1. Introduction 8. Non-Simplicial 2. Simplex Definition Elements 3. Degeneracies of 9. Shape Quality Simplices Visualization 4. Shape Quality of 10. Shape Quality Simplices Equivalence 5. Formulae for Sim- 11. Mesh Quality and plices Optimization 6. Voronoi, Delaunay 12. Size Quality of and Riemann Simplices 7. Shape Quality and 13. Universal Quality Delaunay 14. Conclusions Mesh Quality – p. 158/331
  • 221. Shape Measures and Delaunay Critero Delaunay meshes have several smoothness properties. The Delaunay mesh minimizes the maximum value of all the element circumsphere radii. When the circumsphere center of all simplices of a mesh lie in their respective simplex, then the mesh is a Delaunay mesh. In a Delaunay mesh, the sum of all squared edge lengths weighted by the volume of elements sharing that edge is minimal. Mesh Quality – p. 159/331
  • 222. 3D-Delaunay Mesh and Degeneracy In three dimensions, it is well known that Delaunay meshes can include slivers which are degenerate elements. Why ? How to avoid them ? Mesh Quality – p. 160/331
  • 223. Empty Sphere Criterion of Delaunay The empty sphere criterion of Delaunay is not a shape measure, but it can be used like a shape measure in an edge swapping algorithm. Mesh Quality – p. 161/331
  • 224. Edge Swapping and θmin Shape Measu During edge swapping, using the empty sphere criterion (Delaunay criterion) ⇐⇒ Using the θmin shape measure (maximize the minimum of the angles). θ3 θ3 θ6 θ1 θ2 θ1 θ5 θ4 θ6 θ5 θ2 θ4 Mesh Quality – p. 162/331
  • 225. What to Retain The empty sphere criterion of Delaunay is not a shape measure but it can be used as a shape measure. Mesh Quality – p. 163/331
  • 226. What to Retain The empty sphere criterion of Delaunay is not a shape measure but it can be used as a shape measure. In two dimensions, in the edge swapping algorithm (Lawson’s method), the empty sphere criterion of Delaunay is equivalent to the θmin shape measure. Mesh Quality – p. 163/331
  • 227. What to Retain The empty sphere criterion of Delaunay is not a shape measure but it can be used as a shape measure. In two dimensions, in the edge swapping algorithm (Lawson’s method), the empty sphere criterion of Delaunay is equivalent to the θmin shape measure. There is a multitude of valid shape measures, and thus a multitude of generalizations of the Delaunay mesh. Mesh Quality – p. 163/331
  • 228. Delaunay and Circumscribed Sphere As the circumscribed sphere of a tetrahedron gets larger, there are more chances that another vertex of the mesh happens to be in this sphere, and the chances that this tetrahedron and the mesh satisfy the Delaunay criterion get smaller. As the circumscribed sphere of a tetrahedron gets smaller, there are fewer chances that another vertex of the mesh happens to be in this sphere, and the chances that this tetrahedron and the mesh satisfy the Delaunay criterion get bigger. Mesh Quality – p. 164/331
  • 229. Circumscribed Sphere of Infinite Radi The tetrahedra that degenerate into a fin, into a cap, into a crystal, into a spindle and into a splitter D D D h A h A C A h C h C B B B D D A h C h C A h h B B have a circumscribed sphere of infinite radius. Mesh Quality – p. 165/331
  • 230. Circumscribed Sphere of Bounded Ra The tetrahedra that degenerate into a sliver, into a wedge, into a slat, into a needle and into a Big Crunch D D h h D A C A C h C A h B B B D D h C A hh h C hh h A B h Bh have a circumscribed sphere of bounded radius. Mesh Quality – p. 166/331
  • 231. What to Retain The empty sphere criterion of Delaunay is not a valid shape measure sensitive to all the possible degeneracies of the tetrahedron. Mesh Quality – p. 167/331
  • 232. Circumscribed Sphere of Bounded Ra Amongst the degenerate tetrahedra that have a circumscribed sphere of bounded radius, the wedge, the slat, the needle and the Big Crunch can be eliminitated D h D A C h C A h B B D D h C A hh h C hh h A B h Bh since they have several superimposed vertices. Mesh Quality – p. 168/331
  • 233. The Sliver And so, finally, we come to the sliver, D h C A C A D B B a degenerate tetrahedron having disjoint vertices and a bounded circumscribed sphere radius, which makes it “Delaunay-admissible”. Mesh Quality – p. 169/331
  • 234. Non-Convex Quadrilateral It is forbidden to swap an edge of a non-convex quadrilateral. S3 S3 T1 T1 T2 S2 S2 T2 S1 S4 S1 S4 S3 T1 S2 T2 S1 S4 S1 S4 Mesh Quality – p. 170/331
  • 235. Non-Convex Quadrilateral S3 Two adjacent triangles forming a non-convex quadrilateral necessa- T1 T2 rily satisfy the empty S2 sphere criterion of S1 S4 Delaunay. Mesh Quality – p. 171/331
  • 236. Loss of the Convexity Property in 3D Mesh Quality – p. 172/331
  • 237. What to Retain The empty sphere criterion of Delaunay is more or less a simplicial shape measure. Mesh Quality – p. 173/331
  • 238. What to Retain The empty sphere criterion of Delaunay is more or less a simplicial shape measure. The empty sphere criterion of Delaunay is not sensitive to all the possible degeneracies of the tetrahedron. Mesh Quality – p. 173/331
  • 239. What to Retain The empty sphere criterion of Delaunay is more or less a simplicial shape measure. The empty sphere criterion of Delaunay is not sensitive to all the possible degeneracies of the tetrahedron. A valid shape measure, sensitive to all the possible degeneracies of the tetrahedron, used in an edge swapping and face swapping algorithm should lead to a mesh that is not a Delaunay mesh, but that is of better quality. Mesh Quality – p. 173/331
  • 240. Table of Contents 1. Introduction 8. Non-Simplicial 2. Simplex Definition Elements 3. Degeneracies of 9. Shape Quality Simplices Visualization 4. Shape Quality of 10. Shape Quality Simplices Equivalence 5. Formulae for Sim- 11. Mesh Quality and plices Optimization 6. Voronoi, Delaunay 12. Size Quality of and Riemann Simplices 7. Shape Quality and 13. Universal Quality Delaunay 14. Conclusions Mesh Quality – p. 174/331
  • 241. Non-Simplicial Elements This section proposes a method to generalize the notions of regularity, of degeneration and of shape measure of simplices to non simplicial elements ; i.e., to quadrilaterals in two dimensions, to prisms and hexahedra in three dimensions. Mesh Quality – p. 175/331
  • 242. Non-Simplicial Elements On Element Shape Measures for Mesh Optimization PAUL L ABBÉ , J ULIEN D OMPIERRE , F RANÇOIS G UIBAULT AND R ICARDO C AMARERO Presented at the 2nd Symposium on Trends in Unstructured Mesh Generation, Fifth US National Congress on Computational Mechanics, 4–6 august 1999 University of Colorado at Boulder. Mesh Quality – p. 176/331
  • 243. Regularity Generalization An equilateral quadrilateral, ie that has four edges of same length, is not necessarily a square... Mesh Quality – p. 177/331
  • 244. Regularity Generalization An equilateral quadrilateral, ie that has four edges of same length, is not necessarily a square... Définition : An element, be it simplicial or not, is regular if it maximizes its measure for a given measure of its boundary. Mesh Quality – p. 177/331
  • 245. Regularity Generalization An equilateral quadrilateral, ie that has four edges of same length, is not necessarily a square... Définition : An element, be it simplicial or not, is regular if it maximizes its measure for a given measure of its boundary. The equilateral triangle is regular because it maximizes its area for a given perimiter. Mesh Quality – p. 177/331
  • 246. Regularity Generalization An equilateral quadrilateral, ie that has four edges of same length, is not necessarily a square... Définition : An element, be it simplicial or not, is regular if it maximizes its measure for a given measure of its boundary. The equilateral triangle is regular because it maximizes its area for a given perimiter. The equilateral tetrahedron is regular because it maximizes its volume for a given surface of its faces. Mesh Quality – p. 177/331
  • 247. Regular Non Simplicial Elements The regular quadrilateral is the square. The regular hexahedron is the cube. The regular prism is the ... regular prism ! Its two triangular faces are equilateral triangle whose edges measure a. Its three quadrilateral faces are rectangles that have a base of √ length a and a height of length a/ 3. Mesh Quality – p. 178/331
  • 248. Quality of Non Simplicial Elements Proposed Extension : The shape measure of a non simplicial element is given by the minimum shape measure of the corner simplices constructed from each vertex of the element and of its neighbors. Mesh Quality – p. 179/331
  • 249. Shape Measure of a Quadrilateral The shape measure of a quadrilateral is the minimum of the shape measure of its four corner triangles formed by its four vertices. D C D C D C D C A B A BA B BA Mesh Quality – p. 180/331
  • 250. Shape Measure of a Prism The shape measure of a prism is the minimum of the shape measure of its six corner tetrahedron formed by its six vertices. F D E F C C C D E A B A B A B C F F F A B D E D E D E C A B Mesh Quality – p. 181/331
  • 251. Shape Measure of an Hexahedron The shape measure of an hexahedron is the minimum of its eight corner tetrahedron formed by its eight vertices. G H E F H G D C D C D C E F AHB A B HB A H D C G G G A B E F E F F E C D A B Mesh Quality – p. 182/331
  • 252. Shape of the Corner Simplex The corner simplices constructed for the non simplicial elements are not regular simplices. For the square, the four corner triangles are isosceles right-angled triangles. For the cube, the eight corner tetrahedra are isosceles right-angled tetrahedra. For the regular prism, the six corner tetrahedra are tetrahedron with an equilateral triangle of side a, √ a fourth perpendicular and edge of length a/ 3. Mesh Quality – p. 183/331
  • 253. Shape of the Corner Simplex Each non simplicial shape measure has to be normalized so as to be a shape measure equal to unit value for regular non simplicial elements. ρ η θmin γ √ √ 2 3 3 3 Square √ 1+ 2 2 4 √ 1+ 2 √ √ √ 18√ 1 2 arcsin(1/ 22+12 3) 3 √6 Prism √ 5(7+ 13) √ 3 2 √ 6 arcsin(1/ 3)−π 7+ 13 √ √ √ √ √ 2 3 2 arcsin((2− 2)/(2 3)) Cube 3−1 3 2 √ 6 arcsin(1/ 3)−π 3−1 Mesh Quality – p. 184/331
  • 254. Degenerate Non Simplicial Elements Définition :A non simplicial element is degenerate if at least one of its corner simplices is degenerate. If at least one of the corner simplices is more than degenerate, meaning that it is inverted (of negative norm), then the non simplicial element is concave and is also considered degenerate. Mesh Quality – p. 185/331
  • 255. Twisted Non Simplicial Elements In three dimensions, the definition of the shape measure of non simplicial elements has one flaw : it is not sensitive to twisted elements. E D F E C C C F D A B A B A B C E E E A B F D F D F D C A B Mesh Quality – p. 186/331
  • 256. Twist of Quadrilateral Faces A critera used to measure the twist of a quadrilateral face ABCD is to consider the dihedral angle between the triangles ABC and ACD on one hand, and between the triangles ABD and BCD on the other hand. If these dihedral angles are π, then the quadrilateral face is a plane (not twisted). The twist in the quadrilateral increases as the angles differ from π. Mesh Quality – p. 187/331
  • 257. Twist of Quadrilateral Faces Definition :Given a valid simplicial shape measure, the twist of a quadrilateral face is equal to the value of the shape measure for the tetrahedron constructed by the four vertices of the quadrilateral face. Thus, a plane face has no twist because the four vertices form a degenerated tetrahedron and all valid shape measures are null. As a quadrilateral face is twisted, its vertices move away from coplanarity, and the shape measure of the generated tetrahedron gets larger. Mesh Quality – p. 188/331
  • 258. What to Retain The shape, the degeneration, the convexity, the concavity and the torsion can be rewritten as a function of simplices. An advantage of this approach is that once that the measurement and the shape measures for the simplices are programmed, in Euclidean as well as with a Riemannian metric, the extension for non simplicial elements is direct. Mesh Quality – p. 189/331
  • 259. Table of Contents 1. Introduction 8. Non-Simplicial 2. Simplex Definition Elements 3. Degeneracies of 9. Shape Quality Simplices Visualization 4. Shape Quality of 10. Shape Quality Simplices Equivalence 5. Formulae for Sim- 11. Mesh Quality and plices Optimization 6. Voronoi, Delaunay 12. Size Quality of and Riemann Simplices 7. Shape Quality and 13. Universal Quality Delaunay 14. Conclusions Mesh Quality – p. 190/331
  • 260. Visualizing Shape Measures 1 0.5 QK (C) y 2 C(x, y) x A(0, 1/2) 1 B(0, −1/2) 1 y 0 -1 3 0 -2 0 1 x 2 Position of the three vertices A, B and C of the triangle K used to construct the contour plots of a shape measure. Mesh Quality – p. 191/331