Page 1 The challenge for Numerical Simulation and High Performance Computing at EADS Forum Ter@tec 30.06.2009 - Supelec Yann Barbaux, EADS Directeur Exécutif EADS Innovation Works
Page 1
The challenge for Numerical Simulation and High Performance
Computing at EADS
Forum [email protected] - Supelec
Yann Barbaux, EADSDirecteur Exécutif EADS Innovation Works
Page 2
EADS at a glance…
# 3
# 3
# 1
# 1
Commercial aircraft
Helicopters
Commercial Launchers
Systems of missiles
Satellites
Military transport aircraft
# 1
# 1
Combat aircraft # 1 in Europe
Page 3
EADS Management Structure
Jussi Itävuori
Human Resources
Ralph D. Crosby Jr.
EADS North America
Jean J. Botti
Chief Technical Officer
Marwan Lahoud
Strategy &Marketing
Hans-Peter Ring
Finance
Chairman of the Board Bodo Uebber
Louis Gallois Chief Executive Officer (CEO)
Eurocopter
Lutz Bertling (CEO)
Defence & Security
Stefan Zoller (CEO)
EADS Astrium
François Auque (CEO)
CoordinationEADS Defence & Security
EADS Astrium
François Auque
Airbus
FabriceBrégier (COO)
Tom Enders (CEO)
Airbus MilitaryDomingoUreña-Raso
EADS Innovation Works is the Corporate Research org anization.It is part of the Chief Technical Officer’s organiz ation
Page 4
A testimony from a pioneer…
Page 5
But already at that time, people were looking for simulation and modeling
Early Flight simulator
Simulation of loads
Page 6
Numerical simulation and HPC : summary
Numerical simulation and HPC: what is at stake from an industrial perspective ?
The “Top Ten” challenges
Conclusion
Page 7
EADS : dealing with complex products in a complex environment
The aircraft is a complex system itself, which need s to be optimized through a multi-disciplinary and multi-criteria approach :
Operating in a wide variety of environments :
� Mechanics� Aerodynamics� Systems� Propulsion� On-board energy
� Market expectations (travel conditions, IFE…)
� Performance (range, speed, pax…)� Costs (acquisition, maintenance, fuel…)� Environment (noise, emissions…)� Security & safety
� Air :� Altitude (pressure, temperature)� Icing� Lightning� Winds
� Ground : taxying, storing – atmospheric conditions – EM disturbances…
As part of a complex transportation system :
� Airport infrastructures :� Docking - undocking� Security� Inter-connection / inter-operability
� Air Traffic Management� Communications
Page 8
EADS : dealing with complex products in a complex environment
A complex industrial organization :
� Numerous partners and suppliers
� In numerous sites / geographical locations
� Using different languages
� Using different IT tools
� Facing increased offset demands
Page 9
Numerical simulation: what is at stake ?
• Faster Development through more efficient processes– Top level efficiency & robustness for
numerical simulation
– Step change in design & data processes
– Paradigm shift in design techniques
• Reduced testing time & cost andimproved quality of design– Less testing, fewer wind tunnel models
– Increased use of high Re Testing
• Highly matured product at industrial launching phase
ETW European Transonic Wind TunnelETW European Transonic Wind Tunnel
-15%
-36%
Win
d T
unne
l Tes
t Day
s
2000 2003 2006 2009
Aircraft in development in year
?
-15%
-36%
Win
d T
unne
l Tes
t Day
s
2000 2003 2006 2009
Aircraft in development in year
?
Overall Definition Production
Work Load
Physical Physical Test Test
dominateddominated
Numerical Simulation dominated Save months &
years through integrated simulation
based development
Detailed Definition
Page 10
Limit risks in decision
makingreduce development cycle
Impact of change configuration (up to
certification)
What is at stake ?
Page 11
Design Processes
Analyse & DefineNeeds
DefineTechnical
RequirementsDesignPhysical
Architecture
Integration Processes
Correct
Integrate
Verify
Deliver
Workload 15% Workload 85%
Produce – Re-use - Buy
Integration Processes
Recept
Assemble
Verify
Integration
Validate
Product
Produce – Reuse - Buy
Design Processes
Analyse & DefineNeeds
DefineTechnical
RequirementsVerify
Validate
Evaluate Optimise
DesignFunctional
Architecture
DesignPhysical
Architecture
Workload 60% Workload 40%
Lean in the Right hand side
A change of approach in the V-Cycle
Page 12
The « Top Ten » challenges for NumericalSimulation
� Uncertainties management
� Inverse methods
� Multi-scale modeling
� Domain decomposition
� Coupling of physics
� Techniques for model reduction
� Validation with appropriate test
� Simulation platforms
� Secure distributed simulation
� Virtual certification
Page 13
Tracking Uncertainties or oversize the protections
• Poor knowledge of physical phenomena
• Bad knowledge of input parameters
• Model Uncertainty (3D model or Analytical Formula)
• Numerical Approximations• Investigation of Variability -
Complexity – Diversity• 2 classes of uncertainties in the
literacy :– Epistemic uncertainties– Aleatory uncertainties
Modeling a lightning strike
From Worse case to probability of failure
Page 14
Observed vs Computed Pressure demonstrated the Model relevance Which model for acoustic sources?
KEY POINT = scientific challenge due to complexity of physics•A close cooperation with Astrium experts and top lev el Academics in Math led to modelling and Simulation of Acoustics at lift off
SAVINGS = better knowledge ���� margin reduction• To prepare new configurations of A5: prediction of a full field of pressure around the launcher• To optimize protection: negotiation with ESA and CNES• This “success story” opened an avenue to a lot of optimization topics inside Astrium (and other BUs);
Inverse method to optimize payload protection against over-pressure during Ariane 5 lift off
Page 15
Electromagnetism leads to Multi Scale Modeling
EQUIPMENTS
SYSTEMS
COMPONENTS
Millimeters - µmeters
Centimeters- Millimeters
Meters - Centimeters
ElectroMagnetic Compatibility examples:Emission: from component to systemSusceptibility : from system to component
3D +2D
Power Grounddecoupling
Page 16
Calcul complet
sources moteur + nacelle
Diffraction par l’aile
Domain Decomposition techniques in convective acoustics
Page 17
Nacelle
Flow around the leading edge
Absorbing material conceptCurrent model used to analyze absorbing process
Coupling of physics : modeling of the nacelle implies materials, aerodynamics and acoustic coupling
Page 18
Manufacturing :
LRI Process simulation
Manufacturing :
LRI Process simulation
Engineering :
Mechanical analysisEngineering :
Mechanical analysis
Engineering :
Impact vulnerabilityEngineering :
Impact vulnerability Engineering : Lightning protection
Engineering : Lightning protectionEngineering :
Noise transmissionEngineering :
Noise transmission
Concurrent Analysis DataManagement
Multi disciplinary optimization : weak coupling example
2 levels :
Local=coupling of different physics
Global = concurrent analysis of different local optimizations
Composite Structures
Page 19
−−⋅−+⋅=
)(2
11maxuf 11)1(21
)(Z
x
nfL
xpphx
ω
For A320 and A340 geometry description
Data compression stage 1: From CAD ( 104-105 ) to parametrisation ( 575)
. . .
. . .
. . .
. . .
. . .
In p u t ve c t o r (N n u m b e rs )
O u t p u t ve c t o r (n n u m b e rs )
. . .
C o m p r e s s in g A N N
- i n t e rn a l n o d e
- i n p u t / o u t p u t n o d e
Data compression stage 2 = Non-linear approximation (84 parameters of interest)
•Simplified / surrogate models is a key factor for future innovation;
•Domain of validity assessment for these models is a challenge as coupling optimization is at stake
Model Reduction ���� for architect use
Page 20
Full aircraft
Large component
Small component
Structural detail
Test Model
Mus
care
sear
chfo
cus
Materialcoupon
Virtual testing
Validation with appropriate test
Page 21
Collaborative simulation
Loop process
PALMAS deliverables
Models building
Structure Modeling Architecture/preliminary sizing
FEM Generation
Aerodynamic(refinement of ) external shape
LoadsMass & Aero coupling
Unsteady aeroModal analysis
Masses(Fuel) System
(non-struct mass)installation
Structure sizing
Parametric A/C Geometry
Update of massesvs structure sizing
FEMn
Yes
Sizing/loadsConverged ?
MassesChanged ?
Yes
No
Correlated Masses, Loads and Structure
No
FEM0
Loop process
PALMAS deliverables
Models building
Structure Modeling Architecture/preliminary sizing
FEM Generation
Aerodynamic(refinement of ) external shape
LoadsMass & Aero coupling
Unsteady aeroModal analysis
Masses(Fuel) System
(non-struct mass)installation
Structure sizing
Parametric A/C Geometry
Update of massesvs structure sizing
FEMn
Yes
Sizing/loadsConverged ?
MassesChanged ?
Yes
No
Correlated Masses, Loads and Structure
No
FEM0FEM0
WP2.1: “PALMAS”Loop process
Deliverables
Models building
Performance assessment•Cruise• T/O• etc.
Engine Modeling * Thrust
* Positioning
Aerodynamics* Function of
external shapes
Masses•Fuel vector
• L/G• Payload
• Structure & systems
Aero shapes optim
Parametric A/C Geometry
Correlated Aero-shapes, perfo
Loop process
Deliverables
Models building
Performance assessment•Cruise• T/O• etc.
Engine Modeling * Thrust
* Positioning
Aerodynamics* Function of
external shapes
Masses•Fuel vector
• L/G• Payload
• Structure & systems
Aero shapes optim
Parametric A/C Geometry
Correlated Aero-shapes, perfo
WP2.2: “H/S aero & perfo”
ROBOPT
« common » M&S(parametric concepts)
«co
mm
on»
M&
S(a
ero-
data
)
« common » data
Virtual Plateaux
AeroCity
Trade-offs
Optimization
Page 22
Collaborative simulation
CollaborativeCollaborative
DistributedDistributedworkwork
Data exchanges Data exchanges �������� Model exchangesModel exchangesPLM PLM ((productproduct lifecyclelifecycle Management)Management) �������� SLM SLM (Simulation (Simulation lifecyclelifecycle Management)Management)
Page 23
Security for collaborative M&S, including HPC
Problem– Commercial and Military Security Issues– Legislation– Authorisation and Authentication– Access control to data and services– Multiple security levels, data and applications
Solution (tested or under test)– SIMDAT provides a commercial-grade trust and security infrastructure based on
industrial Web Service specifications
– GRIA contributes dynamic trust (from NextGRID) for relationship management
– NEC’s end-to-end Security infrastructure has been integrated with GRIA for message level security and security context negotiation
– PKI Infrastructure for authentication of users.
– Service manager has control of his service offerings.
– Data ownership and access control.
Page 24
KEY POINT = acceptance (and not scientific)•To make Airworthiness Bodies comfortable with Virtu al certification: more than 10 years of validation & comparisons, demonstration wi th Airbus and suppliers …
SAVINGS = time cycle• To optimize antennas sitting at low cost by simulation• To facilitate and prepare the retrofit or the installation of specific antennas by simulations• To support suppliers with better requirements to limit risks during integration process (later in the cycle)
Dedicated mock-up to perform model validation; High performance computing is compulsory to get confidence
“Virtual Certification”: a milestone in 2006For Antenna Sitting: Green light by DGAC authoritie s
Page 25
Challenges highly linked to HPC performance Challenges highly sensitive today
� Uncertainties management
� Inverse methods
� Multi-scale modeling
� Domain decomposition
� Coupling of physics
� Techniques for model reduction
� Validation with appropriate test
� Simulation platforms
� Secure distributed simulation� Virtual certification
Page 26
Same problematic in « System of Systems approach » in military domain
Preparation of simulation
(Tools)
Performing the simulation
Visualization and post processing
Real time simulations
Support functions:
•Security,
•Repository
•…
Technical repository;:
•Interface spe.,
•Data model
•…
Models environments scenarios
Secured communication
Configuration management
Save, store, capitalize ���� knowledge
Lean in the left hand side
Page 27
… and simulation dedicated to a « virtual » bird, spacecraft …
« hardware in the loop
Preparation of simulation
(Tools)
Performing the simulation
Visualization and post processing
Hardware in the loop
Support functions:
•Security,
•Repository
•…
Technical repository;:
•Interface spe.,
•Data model
•…
Models environments scenarios
Secured communication
Configuration management
Save, store, capitalize ���� knowledge
Page 28
Conclusion 1
M&S can be a leverage to innovation if:
� Certification bodies fully support new methodologies based on Virtualization (� more than 10 years for the “simple” antenna sitting topic)
� Suppliers are closely embedded and can provide relevant model based on our requirements (�optimization of equipment and architecture is at stake): sharing benefits of simulation
� Risk assessment and margin reduction can be addressed by simulation (� methodology, tools and best practices will be brought by other domains than Aeronautics) “from worst case analysis to the probability of feared events”
Page 29
Conclusion 2
Needs for Numerical Simulation are increasing rapidly; needs in the domain of modeling of physics will be fulfilled only if computing costs are “reasonable” (including pre-treatment)EADS needs vary in time / importance depending on Bus & programs, resulting in mostly 2 approaches:
� Outsourced capabilities in the right part of the “V”
� Internal capabilities for the R&D and for the left part of the “V”
� Cloud Computing is an option
Security for distributed simulation in the extended enterprise is a key issue
EADS supports European Initiatives
Page 30
FuSim: Full Flight Simulation – Ease Design Process
• Real-time simulation of maneuver flight of a comple te aircraft – Aerodynamics: Full unsteady Navier-Stokes simulation of flow
– Structures: Full finite element modeling of the airplane
– Flight control: Full simulation of maneuver flight
– Integration: Full interaction of all disciplines
• Loads and stress for the airplane in the whole enve lope– Intelligent analysis of the flight envelope
– Provision of all aerodynamic data for all components and total aircraft
– Prediction of sensitivities – shaping the optimum
• Digital prediction of “Flight Performance” and “Hand ling” prior to first flight– Investigation of flight scenarios and maneuvers
– Full knowledge about aircraft behavior in flight
• Virtual certification prior to production– Full knowledge of product characteristics
– Fully guaranteed prediction accuracy
Page 31
What if Louis Blériot…
Page 32
Had known simulation ?
� TAU Navier-Stokes CFD
� A380 Landing Configuration incl.Landing Gear
Aircraft in Flight
Wind TunnelExperiment
Numerical Flow Simulation
A380 Take-off
TAU Navier-Stokes Computation
DNW Test
Application
Landing Take off
De-rotateApproachγγγγ= -3°°°°
Flare
50 ft
Lift offRotate
35 ft
Lift and drag effects
due to groundGround effect
on tailplane
Ground interferencewith engine jet