getting more from your model with designxplorer€” mesh sizing controls — contact offsets —...
TRANSCRIPT
CAE Associates Inc.
Engineering Consulting Firm in Middlebury, CT specializing in FEA, CFD,
and Electromagnetic analysis.
ANSYS® Channel Partner since 1985 providing sales of the ANSYS®
products, training and technical support.
e-Learning Webinar Series
This presentation is part of a series of e-Learning webinars offered by
CAE Associates.
You can view many of our previous e-Learning session either on our
website or on the CAE Associates YouTube channel:
If you are a New Jersey or New York resident you can earn continuing
education credit for attending the full webinar and completing a survey
which will be emailed to you after the presentation.
CAEA Resource Library
Our Resource Library contains over 250 items including:
— Consulting Case Studies
— Conference and Seminar Presentations
— Software demonstrations
— Useful macros and scripts
The content is searchable and you can download copies of the material to
review at your convenience.
CAEA Engineering Advantage Blog
Our Engineering Advantage Blog offers weekly insights from our
experienced technical staff.
CAEA ANSYS® Training
Classes can be held at our Training Center at CAE Associates or on-site
at your location.
CAE Associates is offering on-line training classes in 2015!
Registration is available on our website.
Why talk about DesignXplorer?
DesignXplorer is a component of ANSYS Workbench that can help you
make your designs more efficient and robust. With DesignXplorer you
can:
— Determine the sensitivity of the systems response to variations in the input
quantities.
— Identify which input variables play a dominant role in the response.
— Develop a surrogate function that enables you to quickly predict the system
output for any parameter combination within the design space.
— Use the surrogate function to determine the optimum input settings for a
defined set of goals and constraints.
— Determine the robustness of the design using probabilistic representations of
the systems input/output relationship.
And by the way, if you are running ANSYS release 18 then you already
have it!
Who Has DesignXplorer?
ANSYS Inc. has bundled the DesignXplorer tool with the following
Mechanical products:
— ANSYS Mechanical Enterprise
— Mechanical Premium
— Mechanical Pro
— Mechanical Enterprise Prep/Post
— DesignSpace and ANSYS AutoDyn-3D
DesignXplorer is also bundled with several of the CFD and Electronics
licenses:
— ANSYS CFD Enterprise, Premium and ANSYS CFD Prep/Post
— ANSYS Mechanical CFD
— ANSYS Academic Associate, Research and Teaching products
— Basic DX core capabilities are now included with ANSYS Electronics Desktop
Parametric Modeling Review
In Parts I and II of this eLearning series we showed you how you can
define input and output parameters using your CAD program, either of the
ANSYS geometry tools, and ANSYS Mechanical.
In case you missed Parts I and II or would like to view them again the
recordings can be found in the resource library at caeai.com or at the CAE
Associates YouTube channel:
Part III–Optimization Using DesignXplorer
The focus of part III of this series is on the optimization tools available with
ANSYS DesignXplorer.
This presentation is also being recorded and will also be available for later
viewing on CAE Associates’ website and YouTube channel.
Additionally CAE Associates is offering a 1-day training class on
DesignXplorer in our Middlebury, CT office on April 7, 2017.
You can register for this class on our website or call our office at (203)758-
2914.
Parametric Modeling Review
The inputs to a parametric ANSYS model can include:
— Geometric dimensions, either from a parametric CAD model or from either of
the ANSYS geometry tools (DesignModeler and SpaceClaim).
— Element properties such as shell thickness and beam cross sections.
— Material coefficients
— Mesh sizing controls
— Contact offsets
— Load magnitudes
Output quantities consist of volume, mass and finite element model result
quantities (reaction forces, deformation, strain, stress, natural frequency,
etc.).
DesignXplorer Tools
DesignXplorer links directly to your parametric model.
DesignXplorer includes several tools that help you evaluate the sensitivity
and robustness of your design:
Optimization Using DesignXplorer
Direct Optimization:
— Uses an iterative approach and the optimum is
determined from actual solutions of the finite element
model.
— The solution process is serial.
— Each optimization run is specific to a single
optimization goal set.
Goal Constraint
DesignXplorer Tools
Parameter Correlation:
— Helps you evaluate the sensitivity levels of each
output parameter to variations of the input.
— Helps you make sure that your optimization study
focusses on the inputs that the system is reactive
to.
— Uses Latin Hypercube sampling to characterize the
design space.
Global Parameter Sensitivities Correlation Scatter
DesignXplorer Tools
Design of Experiments helps you set up a table of design points to
characterize your design space. Custom DOE models can also be
imported into DesignXplorer.
Design points submitted through the ANSYS Remote Solve Manager can
utilize the available computing resources to solve simultaneously.
With an HPC Parametric Pack you can clone all of the licenses needed for
a single design point solution (CAD plugin, Geometry, Mechanical, HPC).
DesignXplorer Tools
Response Surface creates surrogate models
based on the deterministic design point solutions
from the Design of Experiments. The surrogate
models are used to calculate rapid predictive
responses of the system.
— Several Response Surface models are available and
can include automated or manual refinement points
(actual model solves) to increase the accuracy of the
surrogate model.
— The Genetic Aggregate Response Surface
(GARS) minimizes user input sensitivity by
automatically testing and building a best fit model
from the available formulations.
Optimization Using DesignXplorer
Response Surface Optimization:
— The Response Surface surrogate model is
used to determine the optimum designs.
— The Response Surface can be used for
multiple optimization goals without the need to
regenerate the design point solutions (unlike
Direct Optimization).
Response Surface Response Surface Sampling Candidate Designs
DesignXplorer Tools
Six Sigma uses probabilistic distributions of the input to help you
determine if your design is robust.
Input OutputResponse Data
Sample Model
For this webinar we will continue with the Lego Man hip piece model that
we discussed in Parts 1 & 2 of this series.
The goal of our optimization study will be to reduce the amount of material
required while maintaining acceptable part strength.
Sample Model
The geometry inputs represent the dimensions of the part that can be
modified within the design limits.
The dimensions come from the CAD geometry and are as follows:
P11
P12 diameter
P16
P14 Radius
P13 Radius
Sample Model
The structural analysis will assume a symmetric bending load that hip joint
is expected to see when Lego Man’s hip and torso are pressed together.
The output parameters are the part volume, deflection, stress and
predicted life of the part (using the Fatigue Module that is also provided
with the Mechanical Enterprise license).
Demonstration model
The defeatured geometry is attached in Mechanical.
— The total part volume is selected as an output parameter.
Demonstration model
The boundary conditions are:
— ½ symmetry is assumed and applied
using frictionless supports.
— A compression only support on the
peg supports the vertical load.
— A cylindrical support at the end of the
peg is used to constrain the
tangential direction only.
— A remote force with an offset is
applied to the top of the peg.
— Large Deformation is turned on.
Demonstration model
Other output parameters include the total deformation and stress in the
base and peg fillets.
A mesh refinement study of the base fillet mesh was conducted to ensure
that the stress distribution was adequately characterized.
Demonstration model
The CAD geometry was attached in DesignModeler where defeaturing
operations were performed.
The CAD parameters filter out with the prefix were selected in
DesignModeler.
The CAD parameters are then visible in the Parameter Set.
Demonstration model
The Parameter Correlation tool is used to evaluate the strength of the
parameter relationships.
The global sensitivities chart shows a strong correlation of the base
thickness the peg diameter to the local stress in their respective regions.
Note that center thickness is only displaying sensitivity to the volume but
has a negligible effect on deflection and stress.
Demonstration model
Scatter plots also display a strong correlation of the base thickness to the
deflection and stress as indicated by the banded nature of the response
points.
Banded scatter indicating a stronger correlation Wide scatter = weak correlation
Demonstration model
The choice of a Design of Experiments model is an important part of the
Response Surface Optimization.
Certain models allow for additional manual or automatic refinement of the
resulting response surface.
In this example we will be using the Central Composite Design model.
Additional post-DOE
refinement available.
Demonstration model
Design of Experiments Parameter Selection:
— The number of actual solves needed to characterize the design space is
dependent on two things:
• The number of input variables
• The selected DOE model
— The correlation study told us that the back thickness has little affect on the
deflection and stress in the part and thus it is removed from the DOE inputs.
Demonstration model
Design of Experiments Input Parameter setup:
— Parameter input can be defined as either discrete or continuous.
— In this case we will define the geometry input parameters as continues with
upper and lower bounds.
— Note that it is important to test the ranges used prior to the DOE run to be sure
that all of the design points are feasible. Failed design points will negatively
affect the quality of the response surface model.
Demo
The Central Composite Design model with full factorials and four design
variable generates a table of 26 design points.
Demonstration model
The Genetic Aggregation response surface is used to create a surrogate
model from the DOE design point solutions.
The Coefficient of Determination (R2) and the Goodness of Fit are used to
access the quality of the response surface.
Note: with the DOE solution set multiple Response Surface types can be
generated and compared.
Demonstration model
The Optimization method using the Response Surface model to locate
optimum candidates in the design space based on defined objectives and
constraints. Optimum candidates can be evaluated deterministically.
Note: The Response Model can be used multiple times for different
optimization goals. This differs from direct optimization when the design
points solutions are generated with respect to a single goal set.
Optimization Result
When we compare the original design to the optimized design we see the
following:
— A volume reduction of 2.7%.
— The change in the maximum deflection is nominal indicating that the global
part stiffness is unchanged.
— The peak stress is reduced by 33%.
— A fatigue life calculation predicts an increase of the minimum number of cycles
from 156 to over 500,000.
Optimization Result
A 2.7% reduction in volume relates directly to a savings in material cost.
Per hip piece this is not a large number but when you consider that Lego
reported sales of 725 million mini-figures in 2015 the savings in material
cost starts to look pretty attractive.
Imagine if they were to optimize the rest of Lego Man’s parts. That’s bound
to put a smile on someone’s face!