MESH MORPHING BASED SHAPEMESH MORPHING BASED SHAPE OPTIMIZATION OF A CLUTCH LEVER
Abaqus Global User Conference, 2004
Presented by:Ramesh Padmanaban, Ragavendran Vasudevan, , g ,
Radha Krishnan
Detroit Engineered Products, Inc.
Yogesh Wadhera – New Venture Gear
Mike Sheh – Engineous Inc.
Recent Trends in the CAE Industry
Significant improvements in computing hardware:High speed processorsHigh speed processorsNetworked/distributed computing etc.
Significant improvements in Solvers:Significant improvements in Solvers:Reduction in analysis execution timeParallel processing
Availabililty of process integration & automation tools:Ability to automate a routine CAE processy pCreation of automated workflows
Recent Trends in the CAE Industry
Availability of general purpose optimization programs:C d t i l d lti di i li ti i ti t di th tConduct single and multi-discipline optimization studies that utilize the automated workflowsCreate higher level ‘quick turn-around’ response surface modelsmodels
Availability of Morphing & Parametrization Tools:Ability to remove a key bottle neck in the optimization process, i.e, CAD dependencyRapid FE & CFD model generation – in a fraction of time taken b on entional methodstaken by conventional methods
Role of Design Exploration & Optimization
With these trends, the CAE engineer is best positioned for:positioned for:
Conducting DOE studiesConduct Optimization studiesPerform design explorationPerform DFSS (Design for Six Sigma) studies
Clear move from ‘single point design study’ to designClear move from single point design study to design explorationGood positioning of the CAE engineers to truly lead p g g ythe design process
CAE TO LEAD THE DESIGN PROCESS
HARDWARE IMPROVEMENTS
SOLVER SPEED IMPROVEMENTS
CAE not just a validation toolInstead CAE will
IMPROVEMENTS IMPROVEMENTS
PROCESSInstead, CAE will lead the design
process
PROCESS AUTOMATION & OPTIMIZATION TOOLS
MORPHING & PARAMETRIZATION TOOLS
OPTIMIZED
TOOLS TOOLS
DESIGN
Different Shape Optimization ApproachesExplicit definition of nodal movement as design variables:
Rather cumbersome to set upLimited to small shape changes
Shape optimization with CAD in the loop:OPTIMIZER
New values of design parameters
PARAMETRIC CAD SYSTEM
ANALYSIS SOLVERSYSTEM
Regenerated CAD model (new design)
FE PRE-PROCESSOR
Is it OPTIMUM ?
YES
NO
FE PRE PROCESSOR (Auto-meshing)
FE model of the new design
STOP
Different Shape Optimization ApproachesRemeshing from CAD data not fully automatableIf FE model is an assembly comprising of different components and different types of elements remeshingcomponents and different types of elements, remeshing almost impossibleMay be limited to only shape parameters
Mesh Morphing based shape optimization processOPTIMIZER
N l f d iNew values of design parameters
MORPHER (FE/CFD Parametrization tool)
Analysis ready FE model ofAnalysis ready FE model of new design
ANALYSIS SOLVER
Is it OPTIMUM ?NO Is it OPTIMUM ?
STOP
YES
Different Shape Optimization ApproachesIn this process CAD model generation is completely eliminated from the optimization processModels generated by the Morpher are analysis readyProcess inherently robustLarge shape changes are possibleLarge shape changes are possible
Important stages in the Morpher based shape optimization process
Parametrization of the FE (or CFD) modelAnalysis Process AutomationOptimization
Parametrization: High Level Description
Linear Static Analysis Model
Non-linear Static Model
Model
MESHWORKS/MORPHER V3.0
Morphing and
Static Model
Noise & Vib ti d l
Parametrized Model Data BaseMorphing and
Parametrization ToolVibrations model
Crash ModelCrash Model
CFD ModelCFD Model
Types of Design Parameters
SHAPE STRUCTURALFEATURE GENERALSHAPE PARAMETERS
Cross-section of A-pillar
STRUCTURAL PARAMETERS
Thickness of shell structures
FEATURE PARAMETERS
Automated stiffener (bead) creation
GENERAL PARAMETERS
Air bag firing timepillar (stiffness/frequency)
Vent Opening width of Passenger Air Bag
shell structures
Cross-sectional properties of beam members
(bead) creation
Number of stiffener beads and their spacing
Friction coefficient
Column stroke
Stiffness curvePassenger Air Bag (occupant safety)
Front hood angle (external
beam members
Etc.
spacing
Automated punching of holes and slots
Stiffness curve
Etc.
(external aerodynamics)
Rail width & height (crashworthiness)
and slots
Etc.
(crashworthiness)
Tether length & connection location
Etc.
Parametrizing FE/CFD models
Parametrization of the FE/CFD Model/Use Morpher to parametrize existing FE/CFD modelsIntroduce shape, structural, feature and general parametersFE/CFD models become INTELLIGENT PARAMETRIC FE/CFD models
Creating a Design
Design A Design B
Design Design DesignDesign parameter A
Design parameter B
Design parameter C
Morph Set A Morph Set B Morph Set CMorph Set A Morph Set B Morph Set C
Control Zone
Deformation Zone
Fixed Zone
Concept of Parametric FE model (contd.)
DP3
DP1 L th d i i bl
DP2
DP1 – Length design variableDP2 – Height design variableDP3 – Width design variableDP4 – Radius design variable
DP1
DP4
DP1
DP1 – 3.5DP2 – 4 0 Design 1DP2 4.0DP3 – 1.2DP4 – 2.0
Parametrized FE model
g
DP1 – 4.0DP2 – 5.0DP3 – 2.0
Parametrized FE model
Design 2DP4 – 2.5
Optimization loop
DP1 Design 1
Design 2
P met i ed FE model
DP2
Design 3Parametrized FE model g
DP3
Design nDP4 Design nOPTIMUM
Loop 1 Loop 2 Loop 3 Loop n
Analysis Process Automation
Use an analysis process automation tool such as Isight to automate typical analysis process steps executed by the CAEautomate typical analysis process steps executed by the CAE engineerThese steps would typically include:
S b i i f h d l f l i AbSubmission of the model for analysis to AbaqusFrom the results generated, extract the specific output parameters such as maximum deflection, maximum stress, contact pressure etccontact pressure etc.Compute sensitivities of output parameters with respect to design variablesBased on sensitivities generate new values for design variablesBased on sensitivities, generate new values for design variablesExecute Morpher to generate analysis model of new designProceed with the next loop
A l i t ti i ti l f t t dAnalysis process automation is essential for automated optimization process
Optimization Methodology
Optimizationp
FEA Run timeResource & Project calendar timeNumber of Design ParametersNumber of Design ParametersDesign Space – Linear & Non-linear
A t t d R S f M d lAutomated Response Surface Model
Analysis Process Automation & Optimization set up
Meshworks/Morpher
Parametrization & shape change engine
Abaqus analysis
gAnalysis ready FE/CFD model with new shapeiSIGHT (Design
parameter values generation)
q y
NO
FE / CFD Analysis
iSIGHT outputParsing
Results extraction Conve
r-gence
Parsing
optimumYES
Optimization Loop (iSIGHT)
Batch Morpher Shape change
Abaqus Translation
Remote execution of Abaqus
iSIGHT waiting for Abaqus results
Results extraction
Remote deletion of old files
iSIGHT integration with Morpher – Design Parameter file
iSIGHT modifies the values of each design parameter in DP file. A samplefile is shown above. The logical way of automatically modifying the designparameter is using an optimization algorithm.
Results Extraction
iSIGHT extracts the results from the text output file ( from analysis ). A sample file is shown above
Presentation of actual case study – optimization of clutch lever
Objective function : Mi i i D fl ti i th l t h lMinimize Deflection in the clutch lever
Constraints :M i b li i tMaximum stress < baseline maximum stress
Design Variables: Shape of diffe ent feat es of the cl tch le eShape of different features of the clutch lever
Range of the design variables:Range based on manufacturability package space andRange based on manufacturability, package space and element quality constraints
Lever - Loads and BCTwo points arerigidly constrainedg yexcept for the rotationabout y axis
2 86KN
z axis translation is constrained
2.86KN
5.65KN
X axis
Z axis
Material = Cast Iron
Design Variables of the Clutch lever
Thickness of ribsThickness of ribs
Width of webWidth of web
Thickness of webThickness of web
Height of ribsHeight of ribs
Optimization – Design SpaceOptimization – Design Space
Results Review - Mass history plot
X 1000kg
Results Review - Maximum Von Mises stress history plot
MPa
Optimization Statistics
Multi-genetic global optimization scheme usedOptimum obtained after 80 loopsOptimum obtained after 80 loopsTotal completion time for Abaqus analysis, results extraction & morphing in each loop = 30 min.HPJ6700 machine used for analysisAfter obtaining optimum for maximum stiffness, a separate optimization was carried out to minimize massoptimization was carried out to minimize massResults are presented in the subsequent slides
Results of Optimization
Stiffness was improved about 25%Stress levels were maintained below the target of the materialStress levels were maintained below the target of the material yield strengthThere was marginal increase in weight
Results of Optimization for maximizing stiffnessShape change from original to Optimal design
Baseline design
Optimized design
Results of Optimization for minimizing massShape change from original to Optimal design
DV2BaselineOptimumOptimum
DV5 & DV7BaselineOptimumDV4
BaselineOptimum
Optimized design
Baseline design
DV5 & DV7BaselineOptimum
DV9BaselineOptimum
Baseline to Optimum
Conclusions
A robust mesh morphing based shape optimization process has been demonstratedIt can be effectively used on components with complex geometriesSignificant shape change has been effected in a robust manner using this processthis processThe process can be successfully implemented on large system level models with multi-disciplinary constraintsTh ti i d d l b t d t i t lith h (STL)The optimized model can be exported out in stereo-lithography (STL) file format, that can be imported into any CAD system, using which a detailed geometry can be builtU i hi h CAE i l l d h d iUsing this process, the CAE engineer can truly lead the design process