Page 281 -
P. 281

.          .
        264    6. Finite Volume Method
                                         .

                           ..                   This edge has to be
                 .                              removed from the
                            .         .         Delaunay triangulation.


          Figure 6.7. Delaunay triangulation to the Voronoi diagram from Figure 6.5.

          The implication

                   Voronoi diagram  ⇒    Delaunay triangulation,
        which we have just discussed, suggests that we ask about the converse
        statement. We do not want to answer it completely at this point, but we
        give the following sufficient condition.
        Theorem 6.5 If a conforming triangulation of Ω (in the sense of finite
        element methods) consists of nonobtuse triangles exclusively, then it is
        a Delaunay triangulation, and the corresponding Voronoi diagram can be
        constructed by means of the perpendicular bisectors of the triangles’ edges.
        We mention that the centre of the circumcircle of a nonobtuse triangle is
        located within the closure of that triangle.
          In the analysis of the finite volume method, the following relation is
        important.
                                                             , k ∈{1, 2, 3},
        Lemma 6.6 Given a nonobtuse triangle K with vertices a i k
                                                 ∩ K of the control volumes
        then for the corresponding parts Ω i k ,K := Ω i k
           , we have
        Ω i k
                       1               1
                         |K|≤ |Ω i k ,K |≤  |K| ,  k ∈{1, 2, 3} .
                       4               2
        The Donald diagram
        In contrast to the Voronoi diagram, where the construction starts from a
        given point set, the starting point here is a triangulation T h of Ω, which is
        allowed to contain obtuse triangles.
                                                ,k ∈{1, 2, 3}. We define
          Again, let K be a triangle with vertices a i k

                      Ω i k ,K := x ∈ K λ j (x) <λ k (x),j  = k ,

                                                               (cf. (3.51)).
        where λ k denote the barycentric coordinates with respect to a i k
                                              1
          Obviously, the barycentre satisfies a S = (a i 1  + a i 2  + a i 3 ), and (see, for
                                              3
        comparison, Lemma 6.6)
                           3 |Ω i k ,K | = |K| ,  k ∈{1, 2, 3} .     (6.4)
        This relation is a simple consequence of the geometric interpretation of
        the barycentric coordinates as area coordinates given in Section 3.3. The
   276   277   278   279   280   281   282   283   284   285   286