See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228651584

A Mesh-Geometry Based Approach for Mixed-Dimensional Analysis

Article · January 2008

DOI: 10.1007/978-3-540-87921-3_18

CITATIONS READS

3 394

3 authors, including:

Sylvain Bournival Jean-Christophe Cuillière

National Research Council Canada, Boucherville, Canada Université du Québec à Trois-Rivières
6 PUBLICATIONS   40 CITATIONS    61 PUBLICATIONS   588 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Dimensional Metrology of Non-rigid Parts without Specialized Inspection Fixtures View project

CAD CAM CAE integration View project

All content following this page was uploaded by Sylvain Bournival on 07 February 2014.

The user has requested enhancement of the downloaded file.

A Mesh-Geometry Based Approach for
Mixed-Dimensional Analysis

Sylvain Bournival, Jean-Christophe Cuillière, and Vincent François

Université du Québec à Trois-Rivières

3351 Boulevard des Forges, CP 500 Trois Rivières G9A 5H7
Québec, Canada
[email protected], [email protected], [email protected]

Summary. When conducting a finite element analysis, the total number of degrees of freedom
can be dramatically decreased using finite elements such as beams and shells. Because of geo-
metric complexities, entire models (or portions of models) must be meshed using volume ele-
ments in order to obtain accurate simulation results. If however some parts of these models fit
the description of shells or beams, then a mixed-dimensional model containing shell, beam and
volume elements side by side can be used. This approach can significantly reduce the time
needed to mesh and solve the system. Unfortunately, problems arise when trying to connect
elements of different dimensions in part due to the incompatible degrees of freedom and the in-
dependently created meshes. This paper presents a solution to these problems based on the gen-
eration of a compatible mesh composed solely of basic elements and without the requirement of
constraint equations.

Keywords: mixed-dimensional analysis, model reduction, mesh, shell elements, beam elements.

1 Introduction
When conducting a finite element analysis (FEA) the engineer is often confronted with
long calculation time or with a limit imposed on the number of elements. Dimensional
reduction is one of the many techniques employed to reduce the computing time and/or
the number of elements. Using plate, shell or beam elements can considerably decrease
the total number of degrees of freedom (DOF) without loss of accuracy.
However, not all models are suitable for dimensional reduction. Indeed, portions of
a model with a complex geometry must be simulated using volume elements. Some
models, often structures, feature portions of their geometry which can be modeled by
reduced-dimensional elements while other portions must be modeled using volume
elements. Examples of these can be found in [1] and [2]. In these situations it is pos-
sible to use mixed-dimensional models which are models that contain more than one
type of element. It is indeed possible to combine in a model shell and volume ele-
ments for example. The stiffness matrix has no restriction on the type of elements it
can contain.
These mixed-dimensional analyses combine the advantages of each type of ele-
ments, a reduced computing time and decreased number of nodes for 1D and 2D
element types and the ability to mesh complex geometries with volume elements.
300 S. Bournival, J.-C. Cuillière, and V. François

Another advantage is the ability to easily modify the model when evaluating design
scenarios. For example, in an optimization process, if a shell thickness is modified,
only one parameter needs to be changed in a mixed dimensional analysis while, in a
full 3D analysis, the CAD model must be changed.
Two major steps are needed to extract results from a mixed dimensional analysis.
The first one is determining which component of an assembly, or which portion of a
component can be modeled by using reduced dimensional elements. To make this task
simpler and more automated, methods have been developed in order to determine
automatically what can be reduced. The medial axis transform [3-5] and the work
done by Chong in [2] are examples of the progress being made on this subject. Unfor-
tunately, this task usually referred to as geometric idealization is far from being fully
automated and a lot of work remains to be done on this subject. In the work presented
here, idealized models are built by the designer in a CAD system.
Fig. 1 illustrates an example of a mixed dimensional model (or idealized model). A
component is partly modeled by using 3D (volume) elements for complex geometry,
beam elements for long and slender portions of the geometry and shell elements for
portions of the geometry which can be accurately described by a face associated with
a small wall thickness.
The next step in the mixed-dimensional analysis process is to take the idealized
model, mesh it and solve the resulting model to obtain analysis results.
Unfortunately these analyses also face some problems, mainly at the interface be-
tween elements featuring different dimensions. In this paper, a zone where 1D
(beams), 2D (shells) and 3D (volume) elements join is referred to as a dimensional in-
terface. We will demonstrate here how these problems can be overcome solely by
generating a mesh featuring specific arrangements of classical 1D, 2D and 3D finite
elements. This paper focuses on these arrangements when coupling shell elements
with volume elements.

Fig. 1. Example of a mixed dimensional model a) Original model b) Idealized model

2 Problems with Mixed-Dimensional Analysis

When trying to join elements featuring different dimensions, two major problems
arise. The first problem is meeting requirements of mesh continuity across a dimen-
sional interface and the second is incompatible degrees of freedom (DOF) across a
dimensional interface.
 A Mesh-Geometry Based Approach for Mixed-Dimensional Analysis 301

2.1 Mesh Continuity

A mixed dimensional model is created in a CAD system as a group of parts modeled

with various dimensions. The mesh generator sees these parts as different entities and
meshes them independently. This results in an incompatible mesh at dimensional in-
terfaces. Even if the different parts are in contact with each other, if nodes are not co-
incident, these parts are not practically bounded together from a FEA perspective,
which implies that the analysis will fail or that results will be inaccurate.
Problems also arise when only some nodes are coincident while others are not.
Fig. 2 illustrates this specific situation. A W-beam is modeled using three 2D faces.
These faces are then meshed independently with shell elements. Nodes of each face
are not always coincident with other nodes at the interfaces. In the process, coincident
nodes are merged whereas many nodes at the interfaces are left unmerged. Of course,
such a model leads to inaccurate results at the interfaces. High stress concentrations
appear near merged nodes since the model represents a situation where faces in con-
tact are bounded together only at merged nodes locations. We will explain below how
discontinuous binding between parts (with identical or different dimensions) can be
avoided.

Fig. 2. Examples of a mesh continuity problem, coincident nodes are merged while others are
not

2.2 Incompatible Degrees of Freedom

The second major problem faced when using mixed-dimensional analysis is incom-
patible degrees of freedom (DOF) at a dimensional interface. Typical 3D elements
only feature three DOF (3 translations) per node while shell elements typically feature
five or six DOF per node (three translations and two or three rotations).
Consequently, when coupling volume and shell elements, even if nodes of both parts
in contact are merged, problems can still arise in some circumstances. Fig.3 illustrates a
302 S. Bournival, J.-C. Cuillière, and V. François

straight line contact between a shell and a 3D volume. The forces applied on the shell
induce a bending moment in the shell. This bending moment is transmitted trough the
shell thanks to the shell’s rotational DOF. Nodes located at the straight line contact
(common to shell and volume elements) bring about FEA modeling problems as volume
elements do not “follow” rotations of shell element. Assembly of the stiffness matrix as
is leads to a singular matrix or to inconsistent results.
Thus, coincident nodes directly merged this way are akin to ball-joint links. Since
in the FE model, the shell and volume intersect as a series of coincident nodes, the in-
compatibility between DOF results in a situation where the shell and volume are
joined by a hinge.
We will also explain below how this second type of inconsistency at a dimensional
interface can be avoided.

Fig. 3. A shell – volume intersection

3 Alternate solutions
Some techniques were designed to circumvent these problems. Shim [6] describes a
way to couple shell and volume elements by adding constraints equations to the solver.
These equations are based on equating the mechanical work for elements located on
both sides of a dimensional interface and tend to overcome inconsistencies between
DOF. Some FEA software feature functions used to couple shell and volume elements.
These functions add constraints equations akin to those described above, but accuracy
issues related to the use of this type of approaches have been outlined [7].
Another approach to overcome inconsistencies between DOF is to use volume
elements with rotational DOF. Indeed, 3D elements have been designed, based on
Allmans’s triangle with rotational DOF [8]. This triangle, designed for 2D analysis, is
a quadratic element with only three nodes, each with 3 DOF (two translations and one
referred to as a pseudo-rotation). Since then, many tetrahedron element types have
been designed, based on the same principles [9]. Sze [10] has referenced a large num-
ber of such elements. It should be noted however that these elements are designed to
 A Mesh-Geometry Based Approach for Mixed-Dimensional Analysis 303

improve on the classical linear tetrahedral element and the additional rotational DOF
is not designed with the intent of coupling elements of different dimensions. The use
of this type of elements implies that the connection between beams or shells and solid
elements is concentrated at one point or one line. Therefore, stress distributions in
solid parts around these interfaces are completely independent of beam section shapes
and shell thicknesses, which introduce significant inaccuracies.

4 Prior Work
Prior to the work presented here, a method was developed to couple beams and vol-
ume elements using only specific arrangements of typical 1D, and 3D FEA elements
[11]. This method came from an adaptation and combination of two suggestions made
by Craveur [12] and it results in solving both problems mentioned above (mesh conti-
nuity and incompatible DOF) through the two following steps: face splitting and add-
ing mini-beams.
Face splitting aims at solving the mesh continuity problem with the objective of
finding a solution that is independent from both the mesh generator and the solver.
Therefore the best way to force the mesh generator to create nodes at correct locations
is to use a technique which consists in splitting a face located at a dimensional inter-
face, in several pieces. This operation generates one or more faces that fill the space
of the original face. This is done to add at least one vertex at the intersecting point be-
tween the beam and the volume. A standard Delaunay or advancing front unstructured
mesh generator will then automatically place a node at the newly created vertex. In
order to improve accuracy, the shape of face splitting is the same as beam’s section.
This is illustrated in Fig. 4; a W-shaped beam intersects a volume at a specific loca-
tion represented by a point. Two small internal faces (colored in the figure) are cut in-
side the volume’s intersecting face. As it can be seen in Fig. 4, a vertex is located at
the intersecting point with the beam, therefore insuring that the mesh generator will
create a node at that location.
Adding mini-beams aims at solving the problem of incompatible DOF. Beam ele-
ments called mini-beams are placed between the main beam’s node and nearby nodes
on the volume’s boundary. While there is still the ball-joint effect explained earlier,
the fact that these supplementary beams are linked with the main beam solves the
problem. Of course, the mini-beams should not all be collinear (located along a
straight line) which introduces a free rotation around this line. Theoretically, a non-
singular stiffness matrix can be obtained with only a few mini-beams and in the ex-
ample illustrated in fig. 4, this would mean adding only four mini-beams on top of
each triangular edge touching the intersection point. Nevertheless, as will be ex-
plained just below, the distribution of mini-beams has been studied carefully and
optimized in order to obtain analysis results that are as close as possible to the real
connection’s mechanical behavior.
No simple equation could be found to adapt the mini-beams’ elastic properties to
the behavior of the beam’s section at the dimensional interface. Therefore, in an at-
tempt to improve accuracy, a great number of tests have been made to establish the
best possible configuration with the layout of the mini-beams and their properties.
However, no configuration was found to perfectly recreate the behavior in every
circumstance.
304 S. Bournival, J.-C. Cuillière, and V. François

Fig. 4. A beam-volume intersection. Two faces are cut in the original face and mini-beams are
overlaid on top of them.

Fig. 5. Local stress results (Von Mises) for an idealized part with several beam-volume inter-
sections

Nonetheless, an acceptable solution was found when using mini-beams with infi-
nite stiffness. This hypothesis meets the Bernouilli-Euler beam element’s formulation
which states that a beam’s section does not change under stress.
To make the entire beam’s section rigid, mini-beams of infinite stiffness are laid on
every edge of the mesh at the dimensional interface. In the example of fig. 4, this means
that every triangle edge inside the two colored faces (obtained from the splitting face
 A Mesh-Geometry Based Approach for Mixed-Dimensional Analysis 305

process) is overlaid with a mini-beam of infinite stiffness. This effectively translates in

faces of infinite stiffness.
Fig. 5 illustrates a sample stress result on a model composed of four beams inter-
secting at odd angles. The geometry at the intersection of these four beams is modeled
using 3D elements while most of the beams themselves are modeled using beam ele-
ments. Fig. 5 also illustrates how the specific case of hollow beams is processed. In
fact, a specific scheme was developed in order to deal with beams of which neutral
axis does not lie in the material of the beam, which is the case of most hollow beam as
well as channel type beams. A supplementary mini-beam of infinite stiffness is used
to link the node at the beam’s extremity and a random node on the face representing
the beam’s section. This can be observed in fig. 5 at the intersection with the hollow
square section. In this particular case no face splitting is required.

5 Extension to Shell-Volume Intersections

The method proposed in this paper aims at solving the same types of problems (mesh
continuity and incompatibility of DOF) for shell-volume intersections. Thus, the work
presented below is based on an extension of beam-volume intersections procedures to
the case of shell-volume intersections. In the case of shell-volume intersections, di-
mensional interfaces are still processed following the two steps mentioned above
(face splitting and mini-beams generation) but these steps are slightly modified.

5.1 Step One: Face Splitting

The aim of this operation is now to generate a common edge at the intersection be-
tween the shell and the volume. An example is illustrated in fig. 6 where a shell inter-
sects a volume on one of its faces. Face splitting consists in cutting two smaller faces
along the intersection between the shell and the volume (colored in the figure). Like
in the case of beam-volume intersections, the shape of splitting face is consistent with
the shell’s section, which means the shell’s thickness. In the CAD software, two
edges are situated at the dimensional interface, one belonging to the volume (on both
smaller faces) and one belonging to the shell. These edges and their associated verti-
ces are merged to form only one edge. This results in a non-manifold model since this
intersection edge features three adjacent faces.
Once launched, the mesh generator will start by generating nodes on all vertices.
Then it will mesh all edges of the model, including the common edge at the dimen-
sional interface, only once. Then, faces are meshed starting from the edges’ mesh and
therefore the nodes located at the intersection are necessarily coincident between the
volume and the shell.
In some cases this procedure cannot be used directly in order to create a common
edge between the shell and the volume. For example, this is the case when the inter-
section edge between the shell and the volume does not entirely lie on a single face
bounding the volume. An example of this can be seen in fig. 7. A shell intersects a
volume but the intersection edge lies at two different location and extends beyond the
volume.
306 S. Bournival, J.-C. Cuillière, and V. François

Fig. 6. The method applied to a shell-volume intersection. a) 3D Model b) Model after face
splitting operation c) mesh with mini-beams (mini-beams and bounding triangles are repre-
sented using different colors also for clarity).

Fig. 7. In order to obtain a series of common edges, the shell’s face has to be cut several times

In this case, the mesh continuity problem remains even if volume’s intersection
face is split correctly because the shell’s edge at the intersection cannot be merged as
is with an edge on the volume’s boundary.
Since the Application Programming Interface (API) we are using does not feature
any tool to split an edge of a shell, the solution in this case is to split the shell’s face
as well, as illustrated in fig. 7. Splitting the shell introduces two rectangular sub-faces,
which leads to the division of the shell’s intersection edge into four smaller edges.
Two of these edges can now be merged with a coincident edge on the volume’s
 A Mesh-Geometry Based Approach for Mixed-Dimensional Analysis 307

boundary at the intersection. The height of these shell sub-faces is chosen with regard
to the element size as specified in the size map1 imposed in the dimensional interface.
This is done in order to obtain well shaped triangular elements when meshing the
shell, as the automatic mesh generation process is constrained by splitting faces.

5.2 Step Two: Adding Mini-beams

To prevent the hinge effect described earlier (a consequence of incompatible DOF)

mini-beams are also added to both internal faces once the mesh is performed. While
the FE solver would not fail in the case illustrated in fig. 6 as the edge is not a straight
line, adding these mini-beams first models a rigid connection instead of a hinge con-
nection at the intersection curve. Also, like in the case of beam-volume connections, it
models the connection between the shell and the volume more accurately as the
connection is made on a face (representing the shell thickness) instead of a curve, re-
ducing by the way stress concentrations at the interface. The operation is rather
straightforward as the list of all triangular segments belonging to the connecting faces
(contour and inside) can be easily obtained. A mini-beam is then added between both
nodes of each segment so that the entire connecting face is overlaid with mini-beams.
The principle of mini-beams of infinite stiffness used in the case of beam-volume
intersections cannot be used here however as the shell’s section at the interface must
be able to bend. Fig. 8a illustrates this type of specific cases where three forces act to

Fig. 8. Displacements results (Y direction) for a shell-volume intersection. a) 3D model b)

Model with mini-beams of infinite stiffness c) Model with oriented mini-beams d) Illustration
of two mini beam’s orientation.

1
In our work, the size map is a tensorial function describing the desired mesh size at each point
in space.
308 S. Bournival, J.-C. Cuillière, and V. François

bend the shell, but a section of infinite stiffness prevents the shell from deforming
properly (Fig. 8b). In order to solve this problem, oriented mini-beams are used in-
stead of mini-beams with infinite stiffness. Oriented mini-beams are configured with
an infinite moment of inertia in one direction, but with no inertia in the other direction
and a null section area. In the case of fig. 8, this means that the mini-beams have a
moment of inertia Iyy infinitely large (see figure 8d). The stiffness in the other direc-
tions is supplied directly by the shell itself.
Fig. 9 illustrates Von Mises stress results obtained for a simple model where a por-
tion of a shell is modeled using shell elements while the remaining portion uses vol-
ume elements. As it can be seen, the Von Mises stress field is quite continuous across

Fig. 9. Von Mises stress results on a shell (tension). Top face stresses are illustrated for shell
elements.

Fig. 10. Von Mises stress results on a shell (bending). Top face stresses are illustrated for shell
elements. 3D view and side view pointing out the deformed shape (at the bottom).
 A Mesh-Geometry Based Approach for Mixed-Dimensional Analysis 309

the part except for very small discontinuities near the dimensional interface between
volume and shell elements. In this particular case these discontinuities appear because
of the inability to obtain mini-beams of real zero and infinite stiffness. Fig. 10
features the same model under a bending load.

6 Examples
The method described above has been successfully implemented through the design
of a computer program. Practically, the whole process starts from an idealized model,
designed in a commercial CAD system. Technical data such as wall thicknesses (for
shells), section properties and orientation (for beams), boundary conditions and me-
chanical properties are also integrated into the model using an interactive interface, as
well as an imposed size map for mesh generation. After this, the remainder of the
process is fully automated. This automatic process starts with modifications to the
idealized model (essentially face splitting) which are led directly in the CAD system
using an Application Programming Interface (API). This API is also used to provide
the geometric model with technical data mentioned above. This API is first used to
process the identification of intersections between shells and volumes (for what con-
cerns the scope of this paper), and then to apply face splitting algorithms. It is also
used to flag internal faces to be overlaid by the mesh generator with mini-beams and
to merge edges and vertices. Then, the addition of mini-beams is carried out by spe-
cific procedures integrated in the automatic mesh generation system developed by our
research team [13-15].
It can also be mentioned here that the size map constraining mesh generation can
be either specified by the analyst or by an automated process (referred to in our work
as automatic mesh pre-optimization [16-18]). Also, in the context of the work pre-
sented here, the size map is adapted according to the shape of internal faces (resulting
from face splitting).

Table 1. Comparison of the different models

6 legged support Coat hanger Ventilation

gate

Mixed dimensional analysis

Nb. elements 23 171 20 121 42 310
Nb. of DOF 20 295 17 900 47 646
Max displacement (m) 2.85 e-9 2.242 e-9 1.589 e-7
Full 3D model (quadratic tetrahedron)
Nb. elements 88 048 26 608 34 373
Nb. of DOF 410 316 129 201 200 982
Max displacement (m) 2.995 e-9 2.248e-9 1.497 e-7
310 S. Bournival, J.-C. Cuillière, and V. François

Fig. 11. Solid model of the waterslide

Fig. 12. The idealized model after face splitting

Table 1 summarizes a few simple examples modeled with the method described in
this paper. The three different models are compared with results obtained from a full
3D model. The mixed-dimensional models use linear tetrahedrons, linear thin shells
and oriented mini-beams while the 3D models use only quadratic tetrahedron. We
were unable to compare these results with mixed-dimensional models using con-
straints equations as we do not have access to that code.
The use of oriented mini-beams is particularly noticeable in the case of the ventila-
tion gate as the faces at the dimensional interface need to bend in order for the model
to deform correctly. Note that the 3D models are not necessarily more accurate as tet-
rahedron elements are ill suited to model thin shells.
 A Mesh-Geometry Based Approach for Mixed-Dimensional Analysis 311

Fig. 13. Mixed-dimensional mesh of the waterslide

Fig. 14. Von Mises Stress results (on external faces for shell elements)
312 S. Bournival, J.-C. Cuillière, and V. François

The next example illustrates the analysis of a waterslide structure in details. Fig. 12
illustrates the solid CAD model of the waterslide. This model is then idealized and
provided with technical data (boundary conditions, shell thickness and material prop-
erties). For this example, support parts are modeled using 3D elements and the half-
tube using shell elements. Fig. 13 illustrates this model after face splitting. As seen
with the zoom, face splitting results are illustrated using different colors.
Fig 14 illustrates the resulting mixed-dimensional mesh. The 7 volumes and 6 shells
are meshed independently. Oriented mini-beams are then added applying the proper ori-
entation and the mesh is automatically graded around the dimensional interfaces.
The mesh, material properties, boundary conditions and shell properties are finally
exported to a commercial FEA solver. The resulting Von Mises stress field can be ob-
served in fig. 15. From this model, various design scenarios can now be easily
investigated.

7 Conclusions
This paper presents a method intended to couple shell and volume elements for
mixed-dimensional analysis only using specific arrangements of classical 2D and 3D
finite elements. Since the technique is mesh-geometry based it presents the advantage
of being independent from both the mesh generator and the solver. The mesh genera-
tor used in this work is based on advancing front techniques, but a Delaunay mesh
generator could be used as well.
The software implementation of this work has proved that the concepts presented
here are consistent and that the process can be fully automated from an idealized
model added with relevant technical data (shell thicknesses and beam sections proper-
ties, boundary conditions and element size constraints).
In a design optimization context, this method generates reliable FEA results and in
many occasions, the use of this method can provide more accurate results than using
only tetrahedral elements. This is particularly the case when shell thickness is small in
comparison with the size map. Generally, when using only tetrahedral or hexahedral
meshing in the case of mechanical structures with beams and shells, resulting 3D
meshes are either very poor (in terms of elements quality) by the way leading to poor
FEA results, or excessively large (in terms of DOF), if compared to mixed-
dimensional meshes.
Of course, the efficient use of mixed-dimensional analysis is also directly related to
the ability to derive idealized models from 3D parametric and feature-based CAD
models. Consequently, a significant research work is presently ongoing in order to
provide efficient and robust tools for CAD models preparation for FEA and deriving
idealized models is one of the subjects that are presently investigated.

Acknowledgments
This study was carried out as part of a project supported by research funding from
Quebec Nature and Technology Research Fund and by the Natural Sciences and En-
gineering Research Council of Canada (NSERC).
 A Mesh-Geometry Based Approach for Mixed-Dimensional Analysis 313

References
1. Armstrong, C.G., Bridgett, S.J., Donaghy, R.J., Mc Cune, W., McKeag, R.M., Robinson,
D.J.: Techniques for Interactive and Automatic Idealisation of CAD Models. In: Proceed-
ings of the 6th International Conference on Numerical Grid Generation in Computational
Field Simulations, July 1998, pp. 643–662 (1998)
2. Chong, C.S., Kumar, A.S., Lee, K.H.: Automatic solid decomposition and reduction for
non-manifold geometric model generation. Computer Aided Design 36(13), 1357–1369
(2004)
3. Donaghy, R.J., McCune, W., Armstrong, C.G.: Dimensional Reduction of Analysis Mod-
els. In: Proceedings of the 5th International Meshing Roundtable, October 1996, pp. 307–
320 (1996)
4. Monaghan, D.J., Doherty, I.W., McCourt, D., Armstrong, C.G.: Coupling 1D Beams to 3D
Bodies. In: Proceedings of the 7th International Meshing Roundtable, October 1998, pp.
285–293 (1998)
5. Suresh, K.: Generalization of the Kantorovich Method of Dimensional Reduction. In: Pro-
ceedings of the 12th International Meshing Roundtable, September 2003, pp. 261–270
(2003)
6. Shim, K.W.: Mixed Dimensional Coupling in Finite Element Stress Analysis. In: Proceed-
ings of the international meshing roundtable, October 2001, pp. 269–277 (2001)
7. McCune, W., Armstrong, C.G., Robinson, D.J.: Mixed-dimensional coupling in finite ele-
ment models. International Journal for Numerical Methods in Engineering 49(6), 725–750
(2000)
8. Allman, D.J.: A compatible triangular element including vertex rotations for plane elastic-
ity analysis. Computers and Structures 19(1), 1–8 (1984)
9. Pawlak, T.P., Yunus, S.M.: Solid elements with rotational degrees of freedom: Part II - tet-
rahedron elements. International Journal for Numerical Methods in Engineering 31(3),
593–610 (1991)
10. Sze, K.Y., Pan, Y.S.: Hybrid stress tetrahedral elements with Allman’s rotational D.O.F.s.
International Journal for Numerical Methods in Engineering 48(7), 1055–1070 (2000)
11. Bournival, S., Cuillière, J.C., Francois, V.: A Mesh based Method for Coupling 1D and 3D
Finite Elements. In: Proceedings of the 2nd World Congress of Design and Modelling of
Mechanical Systems (CMSM 2007) (March 2007)
12. Craveur, J.C., Marceau, D.: De la CAO au calcul, Paris, Dunod (2001)
13. Cuillière, J.C.: An adaptive method for the automatic triangulation of 3D parametric sur-
faces. Computer Aided Design 30(2), 139–149 (1998)
14. Béchet, E., Cuillière, J.C., Trochu, F.: Generation of a Finite Element Mesh from Stereo-
lithography (STL) Files. Computer Aided design 34(1), 1–17 (2002)
15. Francois, V., Cuillière, J.C.: An a priori adaptive 3D advancing front mesh generator inte-
grated to solid modeling. Recent Advances in Integrated Design and Manufacturing in
Mechanical Engineering, 337–346 (2003)
16. Francois, V., Cuillière, J.C.: Automatic mesh pre-optimization based on the geometric dis-
cretization error. Advances in Engineering Software 31(10), 763–774 (2000)
17. Cuillière, J.C., Maranzana, R.: Automatic and a priori refinement of three-dimensional
meshes based on feature recognition techniques. Advances in Engineering software 30(8),
563–573 (1999)
18. Boulet, A., Cuillière, J.C., Francois, V.: Automatic size map calculations for mesh pre-
optimization in the context of CAD-FEA integration. In: Proceedings of the 2nd World
Congress of Design and Modelling of Mechanical Systems (CMSM 2007) (March 2007)
View publication stats