Scilab Presentation

Scilab Optimization
process control
with mesh morphing
Yann Debray & Hugues-Arthur Garioud – Scilab – ESI Group

2017/10/17
www.esi-group.com 1
Copyright © ESICopyright
Group, 2017.
© ESI
AllGroup,
rights reserved.
2017. All rights reserved. G/EO/17,014
Agenda
Introduction
1. Design of Experiment
2. Model Reduction
3. Optimization
Results & Conclusion
www.esi-group.com 2
Copyright © ESI Group, 2017. All rights reserved. 2

Introduction
OpenFOAM + Scilab = CFD automation
Mesh Perturbations Model reduction Optimization

- Hicks-Henne Sine Bumps - - POD - - Gradient/GA -
Mesh Generation DOE Simulations Validation

- Mesh Morphing - - simpleFoam - - simpleFoam -
DOE Generation Surrogate Modeling Optimization

www.esi-group.com 3
Copyright © ESI Group, 2017. All rights reserved.

I Shape parametrization
Scilab – Hicks & Henne sine bumps
Based on initial airfoil

𝑁
𝑦 = 𝑦𝑏𝑎𝑠𝑖𝑠 + α𝑖 𝑓𝑖 (𝑥)
𝑖=1
log 0.5 𝑡2𝑖

𝑓𝑖 𝑥 = sin π𝑥 log 𝑡1𝑖
One perturbation = One parameter αi

NACA0012 perturbated with 3 sine bumps.
Width t2 = 4 and position t1i = [0.33 0.66 upper 0.66 lower]
www.esi-group.com 4

I Mesh Morphing
OpenFOAM – Mesh Morphing
POD need mesh with same topology
- LaplacianDisplacement -
1] Set template for

PointDisplacement and
DynamicMesh files
2] Copy of the Scilab computed

perturbations
3] Manually adjust with frozen points,

hence avoiding non orthogonal cells
Perturbated NACA0012 mesh and pointDisplacement File
www.esi-group.com 5

I Design Of Experiment
OpenFOAM – DOE simulations
DOE set up
1] From template case,

create new cases
2] Change the
constant/polymesh/points
file
2-level full factorial + center point DOE for 3 parameters varying pressure
www.esi-group.com 6

II Model reduction with POD method
Scilab – Proper Orthogonal Decomposition
4 modes
Modes energy
99.96% of the
global energy
First 4 dominant POD modes for pressure DOE case pressure field projection on 4 modes
POD basis
www.esi-group.com 7

III Optimization process
Scilab - Optimization problem expression
𝐶𝐷
Airfoil Case • Minimize cost function: 𝑓 𝑥 =
𝐶𝐿
• NACA0012
• Considering 3 shape parameters: 𝑥 = (α1 , α2 , α3 )
• Re = 3e6 (Sine bumps amplitude)
• M = 0.15 −0.05 ≤ α1 ≤ 0.05

• Under the constraints: −0.02 ≤ α2 ≤ 0.02
• α=0 −0.02 ≤ α3 ≤ 0.02
And CL > 0
www.esi-group.com 8

Scilab – Optimization cycle
Initialization Minimization following the

New parameter value
Starting parameter x0 gradient
Cost function evaluation
Minimization using
Random population of New population of
selection, mutation and
starting parameters parameters
cross-over
POD field prediction
CD, CL computation
Gradient Method
Cost function evaluation
Genetic Algorithm CYCLE / GENERATION
Cost function gradient
Shared Process evaluation
www.esi-group.com 9

Scilab – Pressure field prediction
Assumption :
Field can be decomposed on the POD basis
𝑁𝑃𝑂𝐷
𝑢 ≃ 𝑎𝑖 φ𝑖
𝑖=1
Projection coefficients (ai) interpolation

with RBF
Field error around airfoil for different prediction methods,

within and out of DOE
www.esi-group.com 10

IV Results: Global Optimum
Within DOE - Linear behavior of the pressure
[0.05 0.02 0.02] - Logical !


IV Results: Reduced model limits
Out of DOE - Low error in lift prediction
[0.07 0.04 0.05] - Model not trained for flow reattachment and recirculation zone

IV Results: Reduced model limits
Out of DOE - High error in pressure drag prediction
[0.07 0.04 0.05] – Iso-pressure on suction side are left-oriented for predicted case (left)

Conclusion
Go further
Optimization Modeling Leveraging other

Scilab capabilities
• 3D/Unsteady/Multiphysic • Model based design
OpenFOAM  Scilab-Xcos • Image & signal processing
• Transonic/Hypersonic
(control of pressure drop) • Reduced order model • Control system
Scilab  OpenFOAM
• Real time optimization for • Statistics
morphing wing
• GUI & automation

T h a n k yo u
yann.debray@esi-group.com

Scilab Presentation

Uploaded by

Copyright:

Available Formats

Scilab Presentation

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Scilab Presentation

Uploaded by

Copyright:

Available Formats

Scilab Optimization

Yann Debray & Hugues-Arthur Garioud – Scilab – ESI Group

Results & Conclusion

Copyright © ESI Group, 2017. All rights reserved. 2

Mesh Perturbations Model reduction Optimization

Mesh Generation DOE Simulations Validation

DOE Generation Surrogate Modeling Optimization

Copyright © ESI Group, 2017. All rights reserved.

Based on initial airfoil

log 0.5 𝑡2𝑖

One perturbation = One parameter αi

Copyright © ESI Group, 2017. All rights reserved.

1] Set template for

2] Copy of the Scilab computed

3] Manually adjust with frozen points,

Copyright © ESI Group, 2017. All rights reserved.

1] From template case,

Copyright © ESI Group, 2017. All rights reserved.

Copyright © ESI Group, 2017. All rights reserved.

• M = 0.15 −0.05 ≤ α1 ≤ 0.05

Copyright © ESI Group, 2017. All rights reserved.

Initialization Minimization following the

Copyright © ESI Group, 2017. All rights reserved.

Projection coefficients (ai) interpolation

Field error around airfoil for different prediction methods,

Copyright © ESI Group, 2017. All rights reserved.

[0.05 0.02 0.02] - Logical !

Copyright © ESI Group, 2017. All rights reserved.

Copyright © ESI Group, 2017. All rights reserved.

Copyright © ESI Group, 2017. All rights reserved.

Optimization Modeling Leveraging other

Copyright © ESI Group, 2017. All rights reserved.

Copyright © ESI Group, 2017. All rights reserved.

You might also like