II What is needed to run a Finite Element Analysis?

What is needed to run a Finite Element Analysis?
This chapter includes material from the book “Practical Finite Element Analysis”. It also has been
reviewed and has additional material added by Matthias Goelke and Jan Grasmannsdorf.
2.1 Basic Information Needed To Run a Finite Element Analysis
In a high level summary, the “working” steps involved in a finite element analysis may be categorized
Modeling (pre-processing)
Visualization of solution results (post-processing)
This image depicts the three elementary working steps involved in a FEM analysis. Some details
about the individual steps are summarized below.
Modeling / Pre-processing
CAD Data
Most commonly any FEM simulation process starts with the import of the component’s (or part’s) CAD
geometry (e.g. CATIA, STEP, UG, IGES, SolidThinking etc.) into the pre-processor i.e. HyperMesh
In many cases, the imported geometry is not ready for meshing. Quite often the geometry needs to
be cleanup first due to
“broken” surfaces
surfaces which are not stitched together
redundant (multiple) surfaces
surfaces which are too small to be meshed in a reasonable way later on
Another issue related to geometry is depicted in the following image:
In the image on the left, the imported geometry is shown. Note the lateral offset of the green edges.
Here, the surface edges (in green) do not meet in a single point i.e. there is a very small lateral offset
of the surface edges. As meshing is carried out with respect to the surfaces, this small offset will be
automatically taken into account during meshing, which, unfortunately will result very poor quality
elements. The image in the middle depicts the meshed “initial” geometry. Note how the mesh is
locally distorted. The updated (cleaned) and meshed geometry is shown on the right.
Here, the surface edges (in green) do not meet in a single point i.e. there is a very small lateral offset
of the surface edges. As meshing is carried out with respect to the surfaces, this small offset will be
automatically taken into account during meshing, which, unfortunately will result very poor quality
Once these “hurdles” are mastered, one needs to ask whether all the CAD information is really
needed. What about little fillets and rounds, tiny holes or even company logos which can often be
found in CAD data? Do they really contribute to the overall performance of the component?
Once the geometry is in an appropriate state, a mesh will be created to approximate the geometry.
Either a beam mesh (1-D), shell mesh (2-D) or a solid mesh (3-D) will be created. This meshing step
is crucial to the finite element analysis as the quality of the mesh directly reflects on the quality
of the results generated. At the same time the number of elements (number of nodes) affects the
computation time. That is the reason why in certain cases a 2D and 1D mesh is preferred over 3D
mesh. For example in sheet metals a 2D approximation of the structure will use much less elements
and thus reduces the CPU time (which is the time while you are desperately waiting for your results)
See the picture above for structures that are typically meshed with 1D, 2D and 3D elements. Which
element type would you choose for which part?
Despite the fact that meshing is (at least optionally) a highly automated process, mesh quality, its
connectivity (i.e. compatibility), and element normals needs to be checked. If necessary, these
element “issues” may need to be improved by updating (altering) the underlying geometry or by
editing single elements.
Material and Property Information
After meshing is completed, material (e.g. Young’s Modulus) and property information (e.g. thickness
values) are assigned to the elements.
Loads, Constraints and Solver Information
Various loads and constraints are added to the model to represent the loading conditions the part(s)
are subjected to. Different load cases can be defined to represent different loading conditions on
the same model. Solver information is also added to tell the solver what kind of analysis is being run,
which results to export, etc.
To determine your relevant loads, your engineering skills are needed. Think of all kinds of load
situations that can occur on your structure and decide whether you want to use them in your
simulation or not. To determine the load from a static or dynamic event, a Multibody Simulation
(MBD) might be helpful.
The FEM model (consisting of nodes, elements, material properties, loads and constraints) is then
exported from within the pre-processor HyperMesh. The exported FEM model, typically called solver
input deck, is an ASCII file based on the specific syntax of the FEM solver chosen for the analysis (e.g.
RADIOSS or OptiStruct). A section out of an OptiStruct solver deck is depicted in the figure below.
As you will see, the bulk of information stored in the analysis file is related to the definition of nodes
(or grids). Each single node is defined via its nodal number (ID) and its x-, y- and z coordinates.
Each element is then in turn via its element number (ID) and its nodes (ID’s are referenced). This
completes the pre-processing phase.
During the solution phase of a simple linear static analysis or an eigenfrequency study, there is not
much for you to do. The default settings of the Finite Element program do handle these classes of
problems pretty well. Practice will show you that if the solution process is aborted by an “error” it
is due to mistakes you have made during the model building phase. Just to mention a few typical
Element quality
Invalid material properties
Material property not assigned to the elements
Insufficiently constrained model (the model shows a rigid body motion due to external
Some of these model issues are discussed and fixed within the “HyperWorks StarterKit Video Series”
Visualization (Post-processing)
Once the solution has ended successfully, post-processing (in HyperView for contour plots and
HyperGraph for 2D/3D plots) of the simulation results is next. Stresses, strains, and deformations are
plotted and examined to see how the part responded to the various loading conditions. Based on
the results, modifications may be made to the part and a new analysis may be run to examine how
the modifications affected the part.
This eventually completes the FEM process.
Practice will show, that in many projects, the above depicted process must be re-entered again,
because simulation results indicate that the part is not performing as requested.
It is quite obvious that going back to CAD (to apply changes) and re-entering the entire FEM process
becomes tedious.
A very efficient (and exciting) technology to speed up this process is called Morphing. Employing
morphing allows the CAE engineer to modify the geometry of the FEM model e.g. change radii,
thickness of ribs, shape of hard corners etc. Quite often the morphed FEM model can be exported
instantaneously (without any remeshing) allowing the CAE engineer to re-run the analysis of the
modified part on the fly.
An example of morphing a given Finite Element model is depicted below:
General remarks
The individual working steps of the FEM process are not only subjected to many “user” errors e.g.
typo while defining material or loads. A lot of attention must be also paid to the chosen modeling
assumptions (for instance, simplification of geometry, chosen element type and size etc.). Even
though the FEM solver may detect some of the most striking errors, the likelihood that your results
are bypassed by “errors” is high.
The following chapters aim at creating awareness about FEM challenges and pitfalls.