View
0
Download
0
Category
Preview:
Citation preview
MSC Software
STRUCTURAL OPTIMIZATION
and its application in
AUTOMOTIVE INDUSTRY
VECOM training
“Simulation methodologies for vehicle structure”
9-10 February 2012, Firenze, Italy
Ing. Mauro Linari
Pro ject Manager
MSC Software S.r. l .
MSC Software
Design M odel
D e s ig n Va r i a b l e s
D e s ig n C o n s t r a i n t s
D e s ig n O b je c t i v e
Finite Element Analysis Program
Numerical Optimization
Program
BRUTE FORCE Method A ‘Brute Force’ coupling of an analysis program and an optimization program suffers from:
‒ Too many design variables
‒ Too many design constraints
‒ Too many detailed analysis
Finite Element Analysis Program
Numerical Optimization Program
Approximate Design Model
Improved Designs
Anal ys is M odel
T h e s t a n d a r d F EM a n d i t s s t r u c t u r a l
r e s p o n s e
• Approximate Design Model concept• It now acts as the interface between
the analysis and the optimizer• Properties of the Approximate Model
‒ Explicit approximations of the objective function and all retained constraints.
‒ It is updated by a full finite element analysis at every design cycle.
Structural OptimizationGeneral Consideration
VECOM 2012 – Firenze Feb 9, 2012
The structural responses which guide the design are implicit functions of the design variables. Function evaluation needs analysis execution at each design modification
Design Optimization is an automatic process of modifying a design to achieve a design objective
while satisfying constraint requirements
MSC Software
• SIZING
– Element and material properties can be considered as
design variables. Also some data in the connection cards
can be design variables
MSC Nastran - Structural OptimizationTypes of structural optimization & supported analysis
• TOPOGRAPHY
• The bead/stamp theory is used to improve the structural
behaviors of a component by varying its shape according
to predefined geometrical constraints
• SHAPE
‒ The shape of the structure can be modified to satisfy the
defined design constraints and to minimize/maximize the
objective function
• TOPOLOGY
‒ The optimal mass distribution in a predefined design
space according to applied loads and constraints.
Mass can be only deleted
• TOPOMETRY
‒ It is a special optimization procedure that uses sizing
design variables to define the optimal mass distribution.
Mass can be locally added or deleted
VECOM 2012 – Firenze Feb 9, 2012
The structural responses of the
following analysis can be directly
used as constraints or objective
functions or to build new responses
by proper equations
‒ STATICS
‒ NORMAL MODES
‒ BUCKLING
‒ FREQUENCY RESPONSE
‒ RANDOM RESPONSE
‒ TRANSIENT RESPONSE
‒ STATIC AEROELASTICITY
‒ FLUTTER
‒ COMPLEX MODES
Dyn
amic
res
po
nse
s d
eriv
ing
fro
m in
tern
al o
r ex
tern
al a
cou
stic
an
alys
is c
an b
e co
nsi
der
ed
MSC Software
MSC Nastran - Structural OptimizationMSC Nastran flow optimization logics
It is repeated each n mathematical optimization design cycles
SIZING & TOPOMETRY
Yes
No
No
Yes
Repeated m types according to user request
SIZING
Mathematical optimization
Many timesNo
Yes
Mode Tracking
if „MODES
SIZING & SHAPE
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
• The basic concept of FSD can be described as follows:
• Fully Stressed Design (FSD) is an alternate approach to the Mathematical Programming (MP) approach for performing automated design tasks.
‒ It resizes the element properties so that each element is at its limit value
under at least one of the applied load conditions.
‒ It produces a „quick-look‟ design with a fraction of the computational cost
and can handle tens of thousand independent design variables.
• The FSD can be used either as a stand-alone tool or as a starting design for an MP run.
‒ User defines the number of FSD cycles (they are executed before the
standard optimization procedure by Mathematical Programming Approach)
• During FSD, the optimizer is not utilized although the standard design model input is used.
MSC Nastran - Sizing OptimizationFully Stressed Design
old
i
allowable
inew
i tt
t = des igned propertyi = index to indicat e which property conta ins the des ig n
parameter and the des ig n response = respons e quant i t y , such as st ress = a rea l number (0 .0 < 1.0)
( ‘old ’ and ‘ new ’ superscr i pt s refer to before and af ter res iz in g)
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
MSC Nastran - Sizing OptimizationFully Stressed Design
VECOM 2012 – Firenze Feb 9, 2012
No Approximate Analysis
MSC Software
• The user specifies a list of „acceptable‟ values for design variables.
‒ Applicable to any design variable.
• The discrete solution is obtained by post-processing the
continuous optimal solution.
• Discrete optimization methods
‒ Design of Experiments (DOE)
• Picks the best discrete design from a set of candidates by evaluating the approximate
objective and constraints. The candidate set is determined using DOE methods.
‒ Conservative Discrete Design (CDD)
• Obtains a conservative solution based on the continuous optimal design by using sensitivity information
‒ Round up
• Simply rounds up to the nearest discrete value
‒ Round off
• Simply rounds off to the nearest discrete value
MSC Nastran - Sizing OptimizationDiscrete variable optimization
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
MSC Nastran - Sizing OptimizationDiscrete variable optimization
VECOM 2012 – Firenze Feb 9, 2012
***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************
(HARD CONVERGENCE ACHIEVED)
NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 5
NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 3
NUMBER OF DISCRETE PROCESSING ANALYSES COMPLETED 1
OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY
---------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
CYCLE APPROXIMATE EXACT OF OF
NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
---------------------------------------------------------------------------------------------------------------
INITIAL 6.614414E+02 1.102773E-01
1 5.662242E+02 5.662401E+02 -2.791767E-05 9.536743E-06
2 5.474170E+02 5.474160E+02 1.783949E-06 2.966779E-03
3 5.474160E+02 5.474160E+02 0.000000E+00 2.966779E-03
3D 5.291842E+02 5.291818E+02 4.613550E-06 3.849362E-02
---------------------------------------------------------------------------------------------------------------
0
DESIGN VARIABLE HISTORY
----------------------------------------------------------------------------------------------------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | DV. ID. | LABEL | INITIAL : 1 : 2 : 3 : 3D : 4 :
----------------------------------------------------------------------------------------------------------------------------------
1 | 1 | X1 | 2.0000E+00 : 1.0000E-01 : 1.0000E-01 : 1.0000E-01 : 1.0000E-01 :
2 | 2 | X2 | 2.0000E+00 : 2.0166E+00 : 2.0266E+00 : 2.0266E+00 : 2.0000E+00 :
3 | 3 | X3 | 2.0000E+00 : 2.8267E+00 : 2.9868E+00 : 2.9868E+00 : 3.0000E+00 :
4 | 4 | X4 | 2.0000E+00 : 1.0000E-01 : 1.0000E-01 : 1.0000E-01 : 1.0000E-01 :
5 | 5 | X5 | 2.0000E+00 : 3.5169E-01 : 1.0000E-01 : 1.0000E-01 : 1.0000E-01 :
6 | 6 | X6 | 2.0000E+00 : 7.8165E-01 : 6.8127E-01 : 6.8127E-01 : 7.0000E-01 :
7 | 7 | X7 | 2.0000E+00 : 1.7342E+00 : 1.6301E+00 : 1.6301E+00 : 1.5000E+00 :
8 | 8 | X8 | 2.0000E+00 : 2.8171E+00 : 2.6749E+00 : 2.6749E+00 : 2.5000E+00 :
Discrete Optimization Cycle Based on previous Standard Optimization Cycle
MSC Software
• Useful when eigenvalues (e.g. first roof bending, first torsional,
etc.) are being designed
• Modes are “tracked” based on a cross-orthogonality check:
where i and i-1 are the eigenvectors relative to two consecutive optimization cycles
‒ If the mode order is not changed after the last optimization cycle the
diagonal term of the resulting matrix is non-null (with corresponding null
out of diagonal terms
‒ If the mode order is changed the off-diagonal term is non null, highlighting
which modes are involved
‒ In this condition the program updates automatically the response entry to
follow the mode to which we are interested
• The user must only define the modes to be tracked
Sizing OptimizationMode Tracking Features
iii
T
i tM 1
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
• Topology optimization determines the optimal shape of a part
• The design variables are the effectiveness of each element
Topology Optimization within the Design Process
FE ModelBasic Topology Result
Results With Manufacturing ConstraintsSmoothed and RemeshedFinal CAD Geometry
CAD Definition of Design Space
Topology OptimizationBrief Introduction
What Can I Use Topology
Optimization For?
• Static Load Path
‒ Multi-subcase
• Frequency Constraint
• Frequency Response
• Multidisciplinary
‒ Static + Modes + Freq
Response + sizing
• Benefits
‒ Used in early design to obtain component designs and shapes
‒ Used to redesign existing components
S p a r e t i r e m o u n t
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
Cyclic Symmetry constraint
• Motivation for Manufacturing Constraints
‒ Topology optimized designs may require major modifications for production or are not producible at all
Casting Extrusion
Without restriction
Design Space
VECOM 2012 – Firenze Feb 9, 2012
Topology OptimizationManufacturability Constraints
• Common issues with unrestricted topology optimization
‒ Thin beams
‒ Cavities which are not achievable by casting or machining process
‒ Tapered sections
‒ Unsymmetric design even when loads, boundary conditions and design space are fully symmetric
‒ Etc.
MSC SoftwareVECOM 2012 – Firenze Feb 9, 2012
Optimization at AUDIOptimization and correlation NVH in the Full Vehicle Analysis
MSC Software
Topology optimization of the threshold profile (extruded Al.)
VECOM 2012 – Firenze Feb 9, 2012
Optimization at AUDIOptimization and correlation NVH in the Full Vehicle Analysis
MSC Software
• Topology can be combined with other types of optimization classes
Design model
• Maximize the 15th natural frequency
• Red: topology design region with
extrusion + mirror symmetry constraints
• Green: sizing (thickness) variables
• Topology mass saving 50% & sizing
mass no change
15th natural frequency
Initial Design 27.316
Topology only 28.174
Combined Topology/Sizing 29.210
Topology only
Combined sizing/topology
Topology optimizationCombined Sizing/Topology optimization
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
• The main objective is to minimize the compliance of the front
mount beam of the engine
The areas highlighted in blue (relative to the constraints regions) should not be touched by the process of topology optimization
DESIGN SPACE
14 loading conditions
relative to different load
combinations acting in the
holes of the connection in
case of the most critical
maneuvers
XYZ Displ. constraints at 5
selected Grids:
- 6. < Δu > + 6.
Topology optimization – Aircraft Engine MountProblem definition
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
• It shows the structure obtained as result of topology optimization
Topology optimization – Aircraft Engine MountAnalysis of the results – Patran Post-Processing
VECOM 2012 – Firenze Feb 9, 2012
Optimized Geometries ‘SMOOTHED’
Solution withoutmanufacturing constraints
Solution in case ofcasting constraints
DESIGN SPACE
MSC Software
• Element-by-element sizing optimization
• Topology/ Topometrycomparison
‒ Topology optimization is a “0” or
“1” discrete element-by-element
optimization methodology
‒ Topology optimization can be
used to decide which element
should be retained and which
element should be discarded
from the design space
‒ On the other hand, topometry
optimization aims to get a
continuous variation of the
designed properties
• Benefits
• It is good to identify critical design
regions
• It is good to locate where to add or
remove material to improve
structural performance.
• Topometry optimization is good for
finding the optimal location of spot
welds. In particular, topometry
optimization is very useful for some
properties that are not supported in
topology optimization; for example,
PDAMP, PELAS, PMASS,
PBUSH, PVISC, PGAP, PACBAR,
and PFAST
Topometry optimizationBrief introduction
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
BIW model correlation /application of topometry optimization
VECOM 2012 – Firenze Feb 9, 2012
Optimization at AUDI Optimization and correlation NVH in the Full Vehicle Analysis
MSC Software
BIW model correlation /application of topology-topometry optimization
Optimization at AUDI Optimization and correlation NVH in the Full Vehicle Analysis
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
BIW model correlation /application of topology-topometry optimization
Optimization at AUDI Optimization and correlation NVH in the Full Vehicle Analysis
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
• The Finite Element model used in the optimization
at Volvo Cars is a fully trimmed sedan model
– The overall model size is about 12M structural DOFs
– The air in the passenger compartment is modelled with solid
elements and coupled to the structure using ACMODL
automatic coupling
• The part of the body set up as design variables is
a BIW except roof and side outer panels.
– The total number of elements in design space is 418148
(number of design variables)
• Eigenfrequencies are solved up to 750 Hz for both
fluid and structure using ACMS
• Case study 1
Optimization of Noise Transfer Functions (NTF)
• Case study 2
Optimization of Road noise level
20
Example trimmed body (TB) model – windows
removed for interior visibility
Example part of a BIW constituting optimisation
set of 418148 shell elements
Optimization at Volvo Car Corporation NVH CAETopometry optimization for noise reduction
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
Optimization set up:
• Minimize the sum of sound pressure in
Driver‟s and front passenger‟s ear position
• Excitation is applied in rear subframe front
mounts in vertical direction
– this is one of the known major transfer paths
for road noise
• Frequency interval is 20-250 Hz
• Constraint is set on the mass to keep it
unchanged
• The thickness in the design space is
allowed to increase a factor of 2 and
reduce to a factor of 0.5 of its original
thickness
• The lowest allowable thickness is 0.6 mm
• This jobs is run both in MSC Nastran 2008
using BIGDOT and in MSC Nastran 2011
using IPOPT and the results are
comparable
Optimization at Volvo Car Corporation NVH CAE
Case study 1 - Optimization of Noise Transfer Functions (NTF)
Results:
• The NTFs after optimization show a
significant reduction compared to the
original
• More that 10 dB reduction is seen at
high peaks, which is very impressive
• Although NTF curves are a bit different,
both BIGDOT and IPOPT show similar
reduction
Lo
we
r is
be
tte
r
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
Post-processing thickness
distribution after design cycle 6
• Red means thicker and blue means
thinner elements and is relative to the
original thickness of the panel
• Significant difference between BIGDOT
and IPOPT in thickness distribution
22
Thickness distribution with MSC Nastran 2008 + BIGDOT
Thickness distribution with MSC Nastran 2011 and IPOPT
Optimization at Volvo Car Corporation NVH CAE
Case study 1 - Optimization of Noise Transfer Functions (NTF)
• Studying this area more in detail
• Both optimizers find similar critical areas
• IPOPT seems to make a more continuous distribution
with less checkerboard effect
• The results from IPOPT are also more symmetric (at
least for this model) which makes more sense since
the loading and response is symmetric
• This is also appreciated by designers since the car
body is normally as symmetric as possible
VECOM 2012 – Firenze Feb 9, 2012
Rear floor panel
MSC Software
Optimization set up:
• Minimize the sum of sound pressure in
Driver‟s and front passenger‟s ear position
• Excitation is applied in all chassis
attachment points in one load case
• The forces have magnitude and phase and
come from measurements and vehicle
simulations in proprietary software
• Frequency interval is 50 to150 Hz and 200
to 250 Hz
• Constraint is set on the mass to keep it
unchanged
• The thickness in the design space is
allowed to increase a factor of 2 and reduce
to a factor of 0.5 of its original value
• The lowest allowable thickness is 0.6 mm
• This jobs is run only in MSC Nastran 2011
using IPOPT23
Optimization at Volvo Car Corporation NVH CAE
Case study 2 - Optimization of Road noise level
Results:
• Significant reduction in the range 70-150 Hz
• Also large reduction at the tyre resonance
at 230 Hz
• The overall reduction in the front seats are
about 2 dBA
• Reduction is also noticed in the rear seat,
even though not included in optimization
• 2 dBA can be the difference between an
average car and a silent car
5 dB
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
Postprocessing element distribution
Optimization at Volvo Car Corporation NVH CAE
Case study 2 - Optimization of Road noise level
VECOM 2012 – Firenze Feb 9, 2012
Red means thicker and blue means thinner elements
and is relative to the original thickness of the panel
• The suggestions are fairly clear and also
rather symmetric which helps in designing
realistic reinforcements
• Suggestions from the optimizer are to increase
gage for floor and C-pillar area
• Reduction in gage can be done in A-pillar,
B-pillar and front area
• Reductions, however, are always a compromise
with other attributes such as safety and
durability
MSC Software
• Also called Bead or Stamp Optimization
• Generate an innovative design proposal for reinforcement bead
pattern with a given allowable bead dimension (minimum bead
width, maximum bed height, and draw angle).
• Topography optimization is particularly powerful for sheet metal
parts
• It is treated as a special shape optimization and built on SOL 200
shape optimization technology
– New algorithms were developed to generate shape design variables and
shape basis vectors automatically based on the user's provided bead
dimension
• Many design variables are generated in the topography optimization
– The adjoint design sensitivity analysis method and large scale optimizer play
key roles in solving topography optimization problems
• All the analysis supported in SOL 200 can be used
Topography OptimizationGeneral Information
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
• The user can provide allowable bead
dimension
• Grids associated to loads and
boundary conditions are skipped
• Remove and/or add designable grids
• Support PSHELL, PCOMP (PCOMPG),
and PSHEAR
• Sizing, shape, topometry, topology
and topography can be combined in
a single job
Non-design elements
MW
MH
ANG
Non-design elements
Design elements
No buffer zone
Buffer zone
Topography OptimizationTopography Optimization Features
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
Maximize 1st Frequency‒ Des ign the base on ly
MW
MHANG
If needed the position of
some grid points in the
design region can be fixed
by:
– Defining the affected region
– Declaring it in the Topology
Optimization control entry
After Optimization
1th frequency: 654 Hz
Before optimization
1th frequency: 582 Hz
Min imum wid th (MW )= 10 .0
Max imum he igh t (MH)= 20 .0
Draw ang le (ANG) = 70 .0
Gr ids descr ib ing the ho les a re f ixed
Topography OptimizationModal Analysis Example
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
Conventional Design‒ Weight=11.39; F1 = 79.5Hz
- Case 1 -
Object ive: Max Freq
Constraint: none
F1=61.0 Hz
Wt = 10.91
13 cycles
Initial Design is ‘Flat’ PanelR=90., Θ=28.6°
Z=30., t=.08
Wtinit=10.78,
F1init=16.8Hz
- Case 2 -
Object ive: Min Wt
Constraint: F1>35Hz
F1 = 36.0 Hz
Wt = 10.81
6 cycles
Topography OptimizationReal life example
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
• Standard Practice (Manual Procedure)
– The model is manually partitioned into
designed and non-designed parts
– The non-designed part is treated as an
external superelement while the designed
part is treated as the residual
– Component Modes Synthesis is applied to
the external superelement and boundary
matrices are stored in the database
– A separate assembly run is performed that
assembles the boundary matrices into the
residual model for solving system solution
– Efficient in terms of CPU time, but tedious
in data manipulation
• AESO automates the tedious manual
partition process by identifying and
partitioning the whole model into the
designed part and non-designed parts,
and performing CMS reduction for non-
designed parts
• Designed part consists of:
• Grid/elements that are designed and constrained
• Grids/elements that are associated with applied loads
• Rigid elements and MPC equations that are touched
by designed grids
• Two separate Nastran runs
• An AESO creation run
• Automatic partition in designed and fixed parts
• Assembly file creation
• Boundary matrices generation
• An AESO assembly run
• The original design task is performed on the reduced
residual structure
Advantages of AESO Technique
Efficiency
- Design task solved in a reduced residual model
- Smaller Residual (10% or less) higher Speed Up
Easy-to-use
- Removes possible user errors
- Minimal knowledge and experiences in superelements
utomatic xternal uperelement ptimization AESO General Information
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
Dynamic Optimization of A car body
Analysis model statistics
‒ 207098 grid points
‒ 1.24 million DOFs
‒ 209079 elementsVehicle model
0
100
200
300
400
500
600
700
1 2 3 4 5 6 7 8 9
Design Cycle
Clo
ck T
ime
(Min
ute
)
Total Time (Single Run)
Total Time (AESO)
utomatic xternal uperelement ptimization AESO sample - Performance
Design task statistics
‒ Design variables: vary the height and width
of some box cross sections of 183 beam
elements
‒ Objective: minimize the structural weight
‒ Constraints: maintaining 1st, 2nd and 3rd
modes above given limits
The final objective by AESO matches that of a single shot run within a relative error of 1%
Speed up is 4+ fold
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
Analysis Domain Design Domain
Update
Displacement Field
Multiple Loading Condition
The optimization process is subdivided in two domains:Analysis Domain and Design Domain
Non linear static analysis is performed in the analysis design
The displacement field is evaluated and the equivalent load sets are derived from it
The equivalent load sets are transmitted to the design domain
Linear static response optimization is performed by using the equivalent static loads as external loads
The design variables are updated in the design domain and non linear static analysis is performed again with the updated design variables
VECOM 2012 – Firenze Feb 9, 2012
Nonlinear Response OptimizationEquivalent Static Loads Method (ESLNRO)
MSC Software
Nonlinear Response OptimizationEquivalent Static Loads Method (ESLNRO)
MSC Software
Topology Optimization with Contact
• What is supported
‒ Sizing, shape and topology optimization
task
‒ Nonlinear static analysis response
(ANALYSIS = NLSTAT)
‒ Geometric and Material Nonlinearities
‒ Response type like displacement (DISP),
STRESS, WEIGHT and VOLUME
‒ Also synthetic responses are supported
(DRESP2)
‒ Contact Conditions and further responses
like reactions (SPCFORCE), fraction of
mass (FRMASS) and compliance (COMP)
Nonlinear Response OptimizationGeneral Behaviors
VECOM 2012 – Firenze Feb 9, 2012
MSC SoftwareVECOM 2012 – Firenze Feb 9, 2012
Nonlinear Response OptimizationImplementation
• The ESLNRO is implemented in MSC
NASTRAN Implicit Nonlinear Solution
(SOL400)
• The single user input file includes
‒ The nonlinear analysis model
‒ The design model with its design variables and
nonlinear response constraints and objective.
• A multiple Nastran invocation strategy
works behind the scenes to harness the
strengths of SOL400 (MSC Nastran
Nonlinear Implicit Solution) and SOL200
(MSC Nastran Optimization Solution)
together to provide an integrated
solution to the design task.
• The submitted job runs to one of the
termination criteria with no need for user
intervention or manual file transfer.
MSC Software
Nonlinear Response OptimizationCase Study – Joined-Wing Aircraft
VECOM 2012 – Firenze Feb 9, 2012
“ The joined wing design is currently being
considered for application to high altitude long
endurance Unmanned Aerial Vehicles (UAVs).”
Consider starboard wing
The length from the wing-tip to the
wing-root is 38 meters and the length
of the chord is 2.5 meters
MSC Software
Nonlinear Response OptimizationCase Study – Joined-Wing Aircraft
Significant nonlinear behavior
– Large displacement
– Critical Physics: snap over buckling behavior at approx 50% load
Normalized Results
– Max Deflection
• Optimized: 1.0, Initial 5.96
– Max VonMises
• Optimized: 1.0, Initial 23.3
O p t i m i z e d S o l u t i o n
0
1000
2000
3000
4000
5000
6000
7000
8000
0 10 20 30
Design Cycle
Weig
ht
- B
lue
0.01
0.10
1.00
10.00
100.00
Max C
on
str
ain
t (
Red
)
(Lo
g S
cale
)
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
Opt imizat ion Process
Optimization using different design modelsMulti-Model Optimization Process
This enables support for the ability to perform design optimization when the design conditions are produced by two or more MD Nastran design models
The M u l t i O p t application initiatesservers that start the processing of theseparate design models up to the pointwhere the optimization is to occur
At this point, a new server is invoked to merge the design information, perform the optimization and partition the results
The servers running the individual models are then resumed in a design loop that is terminated when either convergence is achieved or the maximum design cycles are reached
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
Model 2 – Built for stress AnalysisGrids = 19411Elements = 19640Design Variable = 9
Model 1 – Built for Torsional Stiffness AnalysisGrids = 5613Elements = 5970Design Variables = 10
Model 3 – Built for Acoustic Analysis (FRF)Grids = 13761Structural Elements = 5970Fluid Elements = 7829Design Variables = 10
Each model is built according to the
request of the specific type of analysis
Optimization using different design modelsTest Case using three different input models
Results
The design overcame a violated
constraint while reducing the
weight from 643. to 573.
VECOM 2012 – Firenze Feb 9, 2012
MSC Software
Structural OptimizationOpen Architecture - SCA Framework Overview
• The Simulation Component Architecture (SCA),
enables the delivery of MSC‟s core simulation
technology, as well as user developed
technologies, as reusable software components
• Nastran components wrapped in SCA, are MSC
Nastran functionalities packaged as reusable
services, known as components
MSC Nastran
SCA
…
MSC components
SCA
Inp
ut
File
Pro
cess
or
(IFP
)
SCA
Load
Man
ager
SCASo
luti
on
Man
ager
…
SCA
OU
TPU
T FI
LE P
roce
sso
r
(IFP
)
SCA
Use
r D
efin
ed S
ervi
ce(U
DS1
)
SCA
Use
r D
efin
ed S
ervi
ce(U
DS2
)
…
User components
• SCA framework
utilized CORBA as
the communication
protocol between
services
• „Open‟ Optimization
‒ External Response
‒ OpenMDO
MSC Software
• MSC Insight is an Adams tool originally built for performing parametric
analysis with Adams.
• MSC Insight structure allows to drive Design of Expriment (DoE) of most
popular MSC simulation engines
• In order to investigate :
‒ Parametric sensitivity and coupling (design understanding, test / analysis correlation)
‒ Optimisation strategies (model updating, design improvement)
‒ Robustness of operating points (manufacturing tolerances)
• Related investigations are based on the identification of polynomial
approximations
• After Design of Experiment phase, the same tool can drive a complete
optimisation of the fully non-linear simulation model (instead of using the
polynomial approximation) .
A different approach to OptimizationDoE : the process through MSC Insight
MSC Software
A different approach to OptimizationDoE : the process through MSC Insight
MSC Software
Seal cross section
Body
Door seal has 2 types
of solicitations
Objectives : behavior between
Min/max and mass reduction
A different approach to OptimizationOptimization demonstrator of Car door seal
MSC Software
External thickness =
nominal +/- 0.8 mm
Internal thickness =
nominal +/- 0.5 mm Height =
Nominal +/- 1 mm
A different approach to OptimizationAutomation process and Parametric modeling: Factors (Inputs)
MSC Software
This document is the property of MSC.Software : it can not be copied or given to a 3rd party without MSC.Software written agreement . Page 48
Total mass
Compression
loads
Min / Nom / Max
Horizontal
Radial
A different approach to OptimizationResponses (outputs)
MSC Software
A different approach to OptimizationInvestigations
This document is the property of MSC.Software : it can not be copied or given to a 3rd party without MSC.Software written agreement .
!
Clearance body/door
Compression loads
This values are under min curve
Statistics
Response Surface
Coupling &
robustness
Optimization
MSC SoftwareThis document is the property of MSC.Software : it can not be copied or given to a 3rd party without MSC.Software written agreement .
Result of optimisation
Optimized parameters on response surface
Surface de
réponse
Calcul de
vérification
Marc
Diff %
EP_INT_CENT 1,0 1,0 0,0%
EP_EXT_CENT -0,8 -0,8 0,0%
HAUTEUR_CENT -0,7 -0,7 0,0%
Mass_gram 27,6 27,6 0,0%
Mass_tube_gram 9,9 10,0 -0,6%
H_F_min 6,6 6,6 0,7%
H_F_nom 8,2 8,3 -0,2%
H_F_max 8,8 8,9 -1,0%
H_LSQR_norm 0,39 0,30 27,9%
H_LSQR 3,5 2,9 21,2%
R_F_min 4,6 4,6 1,1%
R_F_nom 6,9 6,7 3,7%
R_F_max 7,7 7,6 1,6%
R_LSQR_norm 0,39 0,33 18,6%
R_LSQR 3,4 3,5 -2,0%
Optimum
A different approach to OptimizationOptimization: starting from DOE
MSC Software
Structural OptimizationConclusions
You can …
Improve performance of existing projects
Combine designers experience to less intuitive
analytical solutions in the preliminary phase of the
project
Take advantage of nonlinear analysis results in design
optimization
Improve efficiency and easy-to-use
Access to in-house/third-party results or optimizers by
taking advantage of the MSC Nastran Open
Architecture
Use stochastic procedure to optimize or check the
robustness of a project
MSC Software
Thanks for your patience
Questions and Answers
Email: mauro.linari@mscsoftware.com
Recommended