31
Efficient Global Optimization Applied to Wind Tunnel Evaluation Based Optimization for Improvement of Flow Control by Plasma Actuator Masahiro Kanazaki(Tokyo Metropolitan University) Takashi Matsuno (Tottori University) Kengo Maeda (Tottori University) Hiromitsu Kawazoe (Tottori University) Japan-Finland Joint Seminar 2013

finland_japan_joint_seminor

Embed Size (px)

DESCRIPTION

Finland Japan Joint Seminor held at Muroran Institute of Technology on 27th June, 2013.

Citation preview

Page 1: finland_japan_joint_seminor

Efficient Global Optimization Applied to Wind Tunnel Evaluation Based Optimization forImprovement of Flow Control by Plasma Actuator

○Masahiro Kanazaki(Tokyo Metropolitan University)Takashi Matsuno (Tottori University)Kengo Maeda (Tottori University)Hiromitsu Kawazoe (Tottori University)

Japan-Finland Joint Seminar 2013

Page 2: finland_japan_joint_seminor

Contents

Introduction Overview of Active Flow Control by Means of Plasma

ActuatorObjectivesOptimization Method Efficient Global Optimization (EGO)Experimental Setup

FormulationResultsConclusions

2

Page 3: finland_japan_joint_seminor

Introduction(1/3)Requirements of flow control around aircraftTake-off and landing Pitching, rolling and yawing motion

➔ Large aerodynamic force underthe large scale flow

3

Complex geometry

Noise

Improvement ofaerodynamics atlanding and take-off

Page 4: finland_japan_joint_seminor

Introduction(2/3)Plasma Actuator: PAElectric device for active flow controlInduced flow (Jet) is appeared by ionization of

the air between exposed electrode and insulated electrodeAlternating current (AC) is supplied.

Small and light weight device

4

Page 5: finland_japan_joint_seminor

Introduction(3/3)Pulse Width Modulation(PWM) PA Efficient AC supplement for PA Optimum values of (T1, T2) or (1/T1, 1/T2) are unknown.

Requirement to find the optimum AC wave formFlow simulation by CFD*: over 10 hours.Real time scale in wind tunnel: 1~ sec.

→ Optimization during a wind tunnel experiment in real time

5

*CFD: Computational Fluid Dynamics

Page 6: finland_japan_joint_seminor

Objectives

Wind tunnel evaluation based optimizationOptimization during a wind tunnel experiment in

real timeEfficient Global Optimization ~ Kriging model based

Genetic Algorithm Improvement of flow control by PA Designing AC wave form

6

Page 7: finland_japan_joint_seminor

7Optimization Method(1/5) Surrogate model:Kriging model

Interpolation based on sampling data Standard error estimation (uncertainty)

)()( iiy xx

global model localized deviationfrom the global model

EI(Expected Improvement) The balance between optimality and uncertainty EI maximum point has possibility to improve the model.

Improvement at a point x is I=max(fmin-Y,0) Expected improvement E[I(x))]=E[max(fmin-Y,0)]To calculate EI,

Jones, D. R., “Efficient Global Optimization of Expensive Black-Box Functions,” J. Glob. Opt., Vol. 13, pp.455-492 1998.

Page 8: finland_japan_joint_seminor

8Optimization Method(2/5)

, :standard distribution, normal density

:standard errors

Surrogate model construction

Multi-objective optimization

and Selection of additional samples

Sampling and Evaluation

Evaluation of additional samples

Termination?

Yes

Knowledge discovery

Knowledge based design

No

Kriging model

Genetic Algorithms

Wind tunnel

Exact

Initial model

Initial designs

Additional designs

Improved model

Image of additional sampling based on EI for minimization problem.

Page 9: finland_japan_joint_seminor

9Optimization Method(3/5) Heuristic search:Genetic algorithm (GA)

Inspired by evolution of life Selection, crossover, mutation

BLX-0.5EI maximization → Multi-modal problem Island GA which divide the population into

subpopulationsMaintain high diversity

Page 10: finland_japan_joint_seminor

Optimization Method(4/5)

Fully automated optimization based on the wind tunnel evaluation.Wind tunnel testing is incorporated into EGO.

• NI LabVIEWTM is employed.

10

Design variable (Power supply)Objective function(Aerodynamic force)

Page 11: finland_japan_joint_seminor

Optimization method(5/5)Flowfield around semicircular cylinder with two PAs Drag minimization by controlling two design

variables related to (T1, T2) Over 1,000 wind tunnel run will be required if full-

factorial design should be carried out.

11

PA off PA on

Page 12: finland_japan_joint_seminor

12Formulation Modulation frequency:

Duty ratio: [%]

m

p

xf

Tf 1

201

1mod

1

2100TTDcycle

Power supply unit provide frequency fp 9kHzand 20/fp as a one unit wave.

[Hz]

Objective function

Design variablesMinimize CD (Drag coefficient)

2 .0 ≤ xm ≤ 90.010.0 ≤ Dcycle ≤ 70.0

Page 13: finland_japan_joint_seminor

13Result(1/5)

Lower xm = Higher jet energy

10 initial samples

Page 14: finland_japan_joint_seminor

14Result(1/5)

Page 15: finland_japan_joint_seminor

15Result(1/5)

Page 16: finland_japan_joint_seminor

16Result(1/5)

Page 17: finland_japan_joint_seminor

17Result(1/5)

Page 18: finland_japan_joint_seminor

18Result(1/5)

Page 19: finland_japan_joint_seminor

19Result(1/5)

Page 20: finland_japan_joint_seminor

20Result(1/5)

Page 21: finland_japan_joint_seminor

21Result(1/5)

Page 22: finland_japan_joint_seminor

22Result(1/5)

Page 23: finland_japan_joint_seminor

23Result(1/5)

Page 24: finland_japan_joint_seminor

24Result(1/5)

Page 25: finland_japan_joint_seminor

25Result(1/5)

Local minimum

Global minimum

After 12 additional sampling

Page 26: finland_japan_joint_seminor

26Result(2/5)

The minimum point could be obtained about 20 wind tunnel runs.

Page 27: finland_japan_joint_seminor

Higher Dcycle can achieve lower CD

Higher Dcycle as DesA provides a higher AC voltage long time to PAs Local optimum DesB can also be found

CD can be also reduced with DesB while the total electrical energy is relatively low. → PAs can control the flow with lower electrical energy under proper PWM driving conditions

27Result(3/5)

DesA

DesB

DesC

Page 28: finland_japan_joint_seminor

28Result(4/5)

x m [-] D cycle [%] f mod [Hz] C D

DesA 2.0 60.0 400.0 0.2985DesB 15.0 25.0 53.3 0.3272DesC 88.0 55.0 9.1 0.4105

DesA

DesB

DesC

Page 29: finland_japan_joint_seminor

29Result(5/5)

x m [-] D cycle [%] f mod [Hz] C D

DesA 2.0 60.0 400.0 0.2985DesB 15.0 25.0 53.3 0.3272DesC 88.0 55.0 9.1 0.4105

DesA

DesB

DesC

DesA: Separated region was reduced, and the streamline was less deformed from the uniform flow

DesB: Separated region was reduced, the streak of smoke far downstream from the model was blurred

Page 30: finland_japan_joint_seminor

ConclusionsWind Tunnel Evaluation–Based OptimizationThe optimization technique successfully

integrated in the operating system of the wind tunnel experiment Automation of the data-acquisition/optimization

processImprovement of Flow Control by Plasma

ActuatorThe cost of optimization based on wind tunnel

evaluation can be drastically reduced Not only global optimum but also local optimum were

found out.

30

Page 31: finland_japan_joint_seminor

31

Kiitos paljon!

Thank you!