Upload
mattia-brenner
View
161
Download
2
Tags:
Embed Size (px)
DESCRIPTION
Overview of the basic concept and most important functionality of CAESES, an upfront CAE system for shape optimization.
Citation preview
DNV GL © 2014 SAFER, SMARTER, GREENER DNV GL © 2014
CAESES
An Upfront CAE System for Parametrically Driven
Shape Optimization
DNV GL © 2014
CAESES
Upfront Optimization
• Post-processing of large sets
• Design explorations
• Formal optimization
• Assessment tools
Upfront CAD
• Simulation-ready
• Variable geometry
• Pre-processing (for CFD)
• Highly automated
CAESES – Upfront CAE System Empowering Simulation
Variable
Geometry
Pre-
processing
Software
Connection
Post-
processing
Optimization &
Assessment
Upfront CFD
• Modern architecture
• Fast, accurate, scalable
• Robust auto meshing
• Batch processing
Mesh
Generation
Flow
Solver
DNV GL © 2014
Process Workflow: Setting up the Automation Chain
Variable
Geometry
Pre-
processing
Software
Connection
Post-
processing
Optimization &
Assessment
DNV GL © 2014
Upfront CAD – Variable Geometry
High-fidelity modeling of complex free
formed surfaces
– Automated variable geometries
– Focus on models subject to internal or external
flows
– In particular, complex surfaces that are
traditionally difficult to parameterize
Simulation-ready CAD
– Based on in-house proprietary CAD kernel
– Right amount of detail at the right time
– Reduced degrees of freedom
Multiple strategies
– Fully-parametric “smart” models
– Partially-parametric models based on morphing
and deformation
– Parametric sensitivities for Adjoint methods
DNV GL © 2014
CAESES Meta-surface Technology
Profile defined using specialized curve types and controlled by user-defined parameters
Initial profile is transformed along a specified path, and it’s parameters are varied based on functional distributions
Proprietary curve-engine and meta-surface technology creates complex surfaces with intelligent parameterization
From 2D thinking to 3D high-fidelity models
DNV GL © 2014
Examples of Functional Parametrics
Propeller blade Radial functions
Compressor Meridional functions
Curved duct Streamwise functions
Ship hull Longitudinal functions Pump volute
Circumferential functions
DNV GL © 2014
1. 2D blade definition • NACA profile • Parameters: chord, thickness, camber, position
2. Radial distributions • 2D section parameters (chord, thickness, etc.) • 3D stacking parameters (pitch, rake, skew)
3. Surface generation • Single blade
Chord Thickness Pitch Rake Skew
x
Example: Propeller
4. Propeller • Parameters: blade number, hub radius
DNV GL © 2014
Example: Propeller – 1 model can create huge number of variants
DNV GL © 2014
Upfront CAD – Smart Modeling
Smart parameters
– Reduced degrees of freedom
(DoF)
– Built-in constraints
– High-fidelity maintained
Examples
– Ship hull (displacement, center of
buoyancy)
– Piston head (fixed compression
volume)
– Volute (specified A/R function)
– Exhaust header (tuning of
relative pipe lengths)
– Flexible tubing (fixed section
circumference)
Standard design: Design and optimize on 2D sections, then stack - Poor quality 3D shape - Many DoF - Takes long time
CAESES smart model: 3D blade designed and optimized directly - High quality blade - Reduced DoF - Fast process
Aeroengine Compressor Blade
Semi-submersible for Offshore Oil Built in constraints:
Pontoon shape variation for sea-keeping - Draft fixed - Displacement fixed * All designs are feasible
DNV GL © 2014
Smart Modeling – Example
10
Semi-submersible for Offshore Oil - Draft fixed - Displacement fixed
DNV GL © 2014
By courtesy of FutureShip
Smart Modeling – Example
Cruise Ship Hull-form Variation - Center of buoyancy fixed - Displacement fixed - Degrees of freedom reduced
DNV GL © 2014
Partially-Parametric Modeling
Variation via curve delta shift
Variation via surface delta shift
Shift and morphing strategies for
transforming existing geometries
Spot shifts for smooth Cartesian &
radial deformations
Delta curve or delta surface shifts
Lackenby shifts for hull-forms
DNV GL © 2014
Optimization using Adjoint methods
Adjoints give shape sensitivity (change of objective function due
to normal displacement of cells)
Traditionally applied to directly morph the shape, but may result
in unusual or infeasible geometry
Instead, use parametric model (fully-/partially-parametric)
Design velocity is the change in normal displacement due to
change in CAD parameters
Parametric sensitivity can be determined from these
Faster and less computationally expensive than direct method
Can consider large design spaces (large number of design
parameters)
Only small displacements are valid
The technology is still in it’s infancy, but we have a novel and
unique approach that we continue to develop
.avg
k
k n
k
kn A
An
n
JJ
With kind permission of: Koenigsegg Automotive AB
DNV GL © 2014
Upfront CAD vs. standard CAD (a summary)
Upfront CAD by CAESES
• Simulation-ready models
• Automatic variation
• For upfront design explorations and optimization
• Specialized modeling – “smart models” with reduced
degrees of freedom
• Right amount of detail at the right time
• Used by the CFD department
Standard CAD for design & production
• Generalized capabilities covering the PLM process
• Manual generation of variants – almost as much
effort for each new design
• Used for detailed design – all the “nuts and bolts”
• Not well suited for simulation
• Effort for de-featuring & making models water-tight
• Used by CAD department
DNV GL © 2014
Upfront CAD – Pre-processing
Simulation-ready geometry
– Watertight
– User specified resolution
– Automatic adaptation for each variant
Exchange file formats
– NURBS surfaces / B-rep
– IGES, STEP, ACIS/SAT
– Discretized geometry STL (multiple
formats)
– Colored STL (e.g STAR-CCM+)
– Extracted colors (e.g. Xflow)
– Specialized formats
– For Numeca geomTurbo
– Propeller free format (PFF)
– Panel meshes for potential codes
DNV GL © 2014
Software Connection – coupling to external CFD codes
Inputs to the external CFD code
Execute CFD in batch
Outputs from the CFD computation
DNV GL © 2014
Software Connection – coupling to external CFD codes
Example: Mixing vessel study with coupling to STAR-CCM+
Geometry:
STL file
Control files:
Java macros, SIM file
Results values:
CSV files
Link to solver
Results files:
Screen shots, field data
DNV GL © 2014
Connecting to external tools
We connect to most commercial and open source codes (as well as in-house codes):
• STAR-CCM+ • FINE/Marine • XFlow
• ANSYS FLUENT • FINE/Turbo • Sesam HydroD
• ANSYS CFX • FINE/Open • NEPTUNE
• ANSYS ICEM • HEXPRESS • Autodesk CFD (CFDesign)
• SHIPFLOW • OpenFOAM • FLOW-3D
• ICON CFD • Snappy Hex Mesh • Pointwise
Contact us to see if your CFD or CAE code can connect with the CAESES.
DNV GL © 2014
Post-Processing
• Result files from the CFD solver
― Images, convergence history, etc.
• Flow-field data can be loaded and
manipulated in CAESES
– Surface plots
– Plane cuts
– Streamlines
– Vector plots
DNV GL © 2014
Upfront Optimization – algorithms
Design engines to vary the geometry
Sobol for design of experiments (DoE)
Local single-objective optimization
Global multi-objective optimization
SSH Resource Manager to configure
and execute distributed or cloud
computing (e.g. HPC)
DNV GL © 2014
Upfront Optimization – assessment
• Design tables – can sort and rank the
variants by selected objective
• Automatic reporting (statistical analysis
such as regressions)
• Spider-web plots for determining
sensitivity to design variables
• Pareto frontier plots
• Direct flow-field comparison of variants
DNV GL © 2014
SAFER, SMARTER, GREENER
www.dnvgl.com
CAESES, Your Upfront CAE System for Shape Optimization
+49-331-96766-0
Design Solve Optimize