« rodin project - École polytechniquemichailidis/rodin_results/sia_proceedings.pdf · « rodin...
TRANSCRIPT
1 RENAULT – Marc ALBERTELLI
« RODIN Project »
Marc Albertelli Research engineer and coordinator of RODIN project
SIA Conference, 18-19 March 2015, France
Session - Optimization
RODIN Robust structural Optimization for Design in INdustry
Accepted within the framework of the 13th FUI call to project
July the 1st 2012 – July the 1st 2015
Labelled by :
With the support from BPI France, CRIF and CGY78
2 RENAULT – Marc ALBERTELLI
Plan
Motivations
Organization
Progress
Design cycle
Topics
Numerical simulation
Applied mathematics
Deliverables
Numerical tools
Methodology
4
production
Intents
Positioning
Target population
Industrial design center
Industries : transport, aeronautic, etc
Industrial expectations
Time savings (productivity)
Mass savings (material cost reduction, CO²)
3 RENAULT – Marc ALBERTELLI
Topology optimization : current process
CAD return Design space
Topology
optimization
+
Set of specifi. stiffness
durability
life cycle,
NVH
manufacturing
etc
Initial
design
Feedbacks
From industrial point of view, the method is attractive for the following reasons:
The execution is straigthforward (compared to parametric approach)
One does not need to create parameters, automate computation workflow, etc
It just requires a design space and the set of specifications
It no longer depends on the designer.
Mass savings : between 3% and 15%
Time savings : achievable but delicate to assess
Despite these advantages, we note that :
Topology Optimization is not systematically used,
The deployment remains modest,
It seems difficult to further enlarge the scope of applications,
And it appears that :
The method is often used for approximately designing simple components,
Engineers struggle to devise robust methodologies dedicated to more complex components,
The solution is just a concept
4 RENAULT – Marc ALBERTELLI
Our (technical) analysis
Threshold=0.1 Threshold=0.5 Threshold=0.9
1) Topology optimization consists in managing the density per element. density 0 the element has virtually diseappeared density 1 the element is full. The optimal shape is defined by the set of full elements.
2) The SIMP approach (or « power law ») clerverly proposes to relax the problem but :
- It does not lead to a clear solution
- There is a lot of intermediate densities that requires to manually adjust where the interfaces lies
- The user has to guess the most probable density threshold
3) The absence of sharp boundary makes difficult the management :
- Of manufacturing constraints (thickness, distance, etc)
- Of surface criteria (pressure, t°, acceleration, etc)
- Of design dependent load,
- Of the export of the solution
Alternative : the level-set method
1) A new way to characterize the shape
Given D, the « design space »,
The shape Ω, is characterized by a level-set function :
The boundary is now well defined. It is given by the iso-zero of the level-set.
The problem of SIMP approach does no more exist.
2) Two other ingredients :
\0
0
0
Dxx
xx
Dxx
ndssjJ ' dIRxtxtxtvxtt
,,0,.,,
5 RENAULT – Marc ALBERTELLI
Plan
Motivations of the projet
Organization
Progress of the project
Main purposes of the project
New process to design massive components Current process
Deliverables
A Topology Optimization tool based on the Level-Set method
An automatic CAD return tool
A new design process
In brief
Topology optimization has proved to be interesting
… but requires some changes to enlarge the scope of applications and to systematize its use
The challenges are : handling manufact. constraints, dealing with true analyses, automat., etc
A new technology is necessary, we choose « level-set method »
designer
simulation Decision maker
CAD return Design space Topology
optimization
+ CDC tenue
fiabilité
acoustique
durée de vie
contraintes de fab.
(fonderie, directe)
performance
etc
6 RENAULT – Marc ALBERTELLI
Some scientific and technical challenges
n°1 : Scale up of Level-set
Machinery level-set
Large scale optimization
Large scale simulation
Multi subcase, ….
n°3 : Analysis
Static linear
Modal analysis, Freq. responses
Material non linearity
Boundary conditions non linearity,
…
n°4 : Pre/post treatment
CAD return
Parametrized CAD
Mesh export
Bulk export, …
CAD return Design space Topology
optimization
+ Set of
specifi. stiffness
durability
life cycle,
NVH
manufacturing
etc
n°2 : Manufacturing constraints
Molding
Maximum and minimum thickness
Minimum distance
Curvature, …
Deliverables
Pre-treatment Optimization Post-treatment
TFC
Visual Env. Topolev (Topol)
Mesh export
Bulk export
CAD return
Visual Env. Bulk Nastran
CAD environment
ESI framework
From Nastran
CAD return Design space Topology
optimization
+ Set of
specifi. stiffness
durability
life cycle,
NVH
manufacturing
etc
7 RENAULT – Marc ALBERTELLI
Academic
Ecole Polytechnique, CMAP
University of Pierre et Marie Curie, labo. JLLions
Inria Bordeaux, Bacchus
Small and Medium Enterprises
Digital Product Simulation
Alneos
Eurodecision
ESI Group
Industrial partners
Renault (coordinator)
AIRBUS Group Innovations
Snecma
Partners
Plan
Motivations of the projet
Organization
Progress
8 RENAULT – Marc ALBERTELLI
1rst Use case
Air cylinder Support (Snecma)
followers
Engine mount (AIRBUS)
Engine mount (Renault) Pilot Use case n°1
...
FreqPropre
min
i
...
1
xxxVonMises
xxxHz
xxxraideur
xxxraideur
MASSEx
CDC
tenue
fiabilité
acoustique
durée de vie
contraintes de fab.
(fonderie, directe)
performance
etc
+ frequency responses
+ maximum thickness
+ Von Mises
+ molding constraints
+ eigenvalue
Solve min mass st stiffnesss constraints
Optimization algortithm dedicated to level-set approach
« level-set » machinery
Airbus and Snecma use-cases
Topolev result
SIMP result
Engine mount (1/2)
Impact of the threshold density
9 RENAULT – Marc ALBERTELLI
Engine mount (2/2)
Unambiguous solution, sharp boundary !
With molding constraints Without molding constraints
Extra use-cases (1/3)
10 RENAULT – Marc ALBERTELLI
Molding (2/3)
Maximum thickness (3/3)
Without max thickness control
11 RENAULT – Marc ALBERTELLI
Complete process
CAD return Geometry
optimization
Topology
optimization
Thank you
For your attention