White Paper

Simulation-Driven Design: Solving the Geometry Problem

Abstract
Over the years, computer-aided design (CAD) system geometry has evolved: wireframes to surfaces to solids, and
parametric to direct. Simulation, however, remains stuck using the same old beam, shell, or simplified-solids paradigm
invented more than 50 years ago. As CAD geometric completeness has grown, so has the effort to convert it into a simpler
form suitable for meshing and analysis. Traditional finite element analysis (FEA) vendors have focused on this task, but this
has just increased the learning curve. The bottom line is that these “CAD to Mesh” steps require many judgment calls, are
labor-intensive and error-prone, and require experts in both simulation and CAD. The result is that broad-based simulation
today is limited to large companies with complex and expensive workflows and, even then, is rarely well integrated within
the conceptual product design process. Our goal with Altair SimSolidTM is to change that. We are pioneering new methods
that work directly on fully featured CAD assemblies and do not create a mesh. With this, you can work in step with your
design process to analyze quickly and efficiently the original CAD geometry without modification or simplification. The
result is that SimSolid is capable of analyzing large assemblies and complex parts that would not be considered practical
with mainstream FEA. In this white paper, we will explain what SimSolid is and how it works and will show examples of
where it can be used. In addition, we will discuss the technological foundations of SimSolid and compare it to methods
used in traditional FEA.

The Geometry Problem

The geometry problem, simply put, is that the geometry of CAD and traditional FEA are different. CAD creates geometry to
define design and manufacturing requirements. FEA needs to transform this into a simplified form to define the mesh. This
disparity between CAD and FEA geometry models needlessly complicates the analysis user paradigm in many areas—some
obvious, some subtle. The obvious complication is the need to dramatically simplify the geometry so that a mesh can be
reliably created. Many decisions must be made as to which parts and part features can be removed without modifying the
geometric design intent. This is an expertise-extensive process. Different people will likely create different results. The
less-than-obvious complication involves the small tweaks to geometry often required to get the traditional FEA mesh
generator to create adequate shaped elements or the special elements and special mesh transitions required to create
connections between parts—bolts, welds, etc. The subtlest complication is the tolerance settings that might need to be
made in the solution methods to account for numeric instabilities caused by poorly shaped meshes, especially in nonlinear
analysis. The geometry problem is the primary inhibitor blocking wider adoption of simulation in design workflows.
Incremental user interface improvement in geometry and meshing is not the solution. A fundamental process change is
required.

The Altair SimSolid Solution

SimSolid takes a different approach. It solves the geometry problem by replacing altogether the underlying core FEA
solution technology. Here are some important attributes of the SimSolid solution:

• The SimSolid methodology analyzes directly the fully featured CAD assembly. Time-consuming model
simplification techniques, such as defeaturing and mid-surfacing, are not required.
• The solution methods used in SimSolid are meshless. There is never a user requirement to create a mesh.
 • The SimSolid methodology is fast and efficient. It provides superior performance metrics for computational time
and memory footprints that allow very large and/or complex assemblies to be solved quickly on desktop-class
PCs.
• The SimSolid method is accurate. SimSolid controls solution accuracy using multipass adaptive analysis.
Adaptivity can be defined on a global or part-local basis. And adaptivity is always active.

All of this is packaged as a lightweight (33 MB) Windows application, which provides both a direct connection to
Onshape cloud documents and a convenient STL interface for the structural simulation of 3Dprinted parts.

The Benefits
SimSolid eliminates the two most time-consuming and expertise extensive tasks of geometry simplification and meshing.
These two steps typically take between 30% and 70% of the total modeling and analysis time, so this represents significant
process improvement. More important, these tasks represent the bulk of the training requirement for traditional FEA. Not
only is less time required but also less training means that a larger pool of users can take advantage of the benefits that
design simulation provides.

The second benefit is expanding the possibilities of what is practical to solve. Instead of reducing the assembly to one part
or a small context of a few parts, a more complete assembly can be solved, simplifying the model setup and load and
constraint specification. Many models that are not practical to use with traditional FEA can be solved using SimSolid.

Here are a few examples of the benefits of using SimSolid: Elimination of part-geometry simplification. With traditional FEA,
the full-fidelity model must first be simplified by examining and then removing small features. Here is an example of the
original CAD geometry and one possible simplification used in traditional FEA. Note that different analysts will likely
simplify the geometry in different ways, further complicating verification of the analysis model.
With SimSolid, all features are left in the model and the full fidelity geometry is used in the analysis. This geometry is a
complex single part of 1,200+ faces and 150+ small holes. Solution time including model setup is less than four minutes,
and reanalysis is fast—less than one minute.

Elimination of part-geometry cleanup. Typical CAD geometry often has small or thin faces and odd face intersections.
SimSolid is indifferent to these types of geometries. Here is a model with many small faces and even an imprint of letters
on the surface. With traditional FEA, many surfaces would need to be edited, merged, or removed altogether.
White Paper

With SimSolid, this geometry can be analyzed directly, without any modification by the user.

Direct simulation of large assemblies. Large assemblies with parts of varying size and shape can be directly analyzed in
SimSolid. Here is an access platform consisting of 153 separate parts. Some parts are solid (ladder rungs), thin and hollow
(tubular frame), complex (floor grate), or small (bolts, nuts, and washers). All geometric details, including threads in the
nuts and bolts, are retained.
The model was imported directly from CAD without modification, and all connections were automatically determined. No
meshing was required. The base was fixed and a simple side load applied. Total solution time (import, setup, and solution)
was less than nine minutes using a desktop PC. A solution of this type would not be practical with traditional FEA.
SimSolid Technology
The SimSolid technology approximates the solution domain using complicated polynomial and non-polynomial functions. It
is an alternative to conventional FEA, where the approximations are built from primitive interpolation polynomials confined
to finite elements of generic shapes like brick, tetrahedral, and wedge. Classes of applicable functions are significantly
extended to accommodate solution-specific functions that, a priori, meet certain solution features, such as
incompressibility conditions, equilibrium equations, and asymptotic analyses around special geometric features. The
accommodation of smart basis functions was made possible due to breakthrough extensions to the mathematical theory of
external approximations, which decouples the basis functions from the underlying geometric shapes. The final functions
are built on the fly from generic sets during the solution sequence. Generic sets are always complete, and their
approximation properties are preserved in all transformations. This property of the basis functions enables the
development of adaptivity strategies that can refine the generic sets in local regions as required to increase solution
accuracy. In SimSolid, an extension of the fundamental concepts of numerical methods consists of the redefinition of the
basic ideas of degrees of freedom (DOF). SimSolid does not use the pointwise DOF inherent in traditional FEA. The SimSolid
DOF are functionals with geometric support in form of volumes, areas, line clouds, and point clouds. This provides the
ability to handle geometric imperfections as well as assembly contact imperfections, like gaps, penetrations, and ragged
contact areas.

Sources of Modeling Error in Traditional FEA

The implementation of analysis in the design process means that analysis results are used to make design decisions. It is,
therefore, important that analysis tools provide results with predictable accuracy. Analysis results validation is a complex
problem because all numerical methods are approximate, and there can be many sources of errors including the major
ones: modeling errors and approximation errors. Modeling errors occur when the CAD geometry model is being modified
to make it suitable for traditional FEA meshing. The modification can include many steps, such as assembly simplification,
part defeaturing, surface idealization, and geometry-face cleanup. Successful meshing is a prerequisite for obtaining any
results in FEA. Even if only global displacements are of interest, the geometry still has to be meshed to the smallest detail.
Furthermore, meshing has to use the correct element type, show the correct element shape (no degeneration or bad aspect
ratio), and have enough elements to model the expected stress pattern. These quality requirements are difficult to satisfy
for complex parts. Adaptive remeshing to satisfy numerical convergence is possible but not practical in many situations
and is not commonly performed in design analysis. For assemblies, the situation gets even worse because meshes in
contact areas of parts must be either compatible or good enough to provide meaningful results. The latter is practically
impossible in case of multiscale assemblies, when large parts are connected through small parts, such as bolts, nuts,
rivets, and pins.
Assembly idealization is also dependent on the solution method. Here is an example of small parts that can be simply
removed in static analysis but that need to be replaced by mass points or artificial bodies with six inertia moments in
dynamic analysis in order not to change mass distribution in the structure. Other sources of error in traditional FEA include
special element consideration for connections. Bolts and welds are problematic in that both special elements and special
mesh patterns are required to model them adequately. To the left are examples of connection idealizations of a bolt
replaced by beam and spider rods.

The final stumbling block is obtaining the solution. Even if model has been successfully meshed, finding the solution is
still not assured. Having meshed complex geometry, the model is often found to be too large to be solved within a
reasonable time or it contains poorly shaped elements that cause instabilities in the numeric of traditional FEA solver
methods. Using the traditional FEA workflow to manage these potential error sources is complex. Training—and
retraining—can be expensive and time-consuming. Occasional (infrequent) use of simulation is especially problematic.
Errors introduced by the misapplication of a user interface workflow are far too common.

Sources of Modeling Error in Traditional FEA

SimSolid is new but it has been extensively tested by both Altair and outside companies in a variety of industries. We have
produced a validation manual, available at www.altairhyperworks.com, that includes tests run as part of our standard
quality-assurance process. One example from the validation manual—plate with a hole under extensional load—is included
here.
Conclusion
For simulation to truly drive the design process, it needs to work in step with each geometric concept and concept
modification. The complexity of traditional FEA eliminates its use in all but the most trivial of design conditions. Simulation
working directly on design geometry provides a path to quick, meaningful answers that can guide designers and engineers
to more optimal design scenarios. Only SimSolid can provide this, by not only eliminating time-consuming and expertise-
extensive geometry-simplification techniques, such as defeaturing and mid-surfacing, but also eliminating the mesh
altogether. The result is a simulation tool that is both:

• fast enough, with respect to both model and solving time, to be used every day, and

• simple enough to be used occasionally without the need for extensive training and monitoring.

Try SimSolid for yourself. We think you will agree that it is how design simulation should be done. For more information
and to conduct a trial simulation of our product, please visit our website at www.altair.com/simsolid.