copyright (c) 2014 intelligent light all rights reservedauthor epd created date 8/6/2014 1:29:44 pm


Using XDB workflows to analyze the2nd AIAA CFD high lift prediction workshop simulations

Earl P.N. DuqueManager of Applied ResearchIntelligent light


Outline• Purpose

– IL’s contribution to the knowledge base

– Get current with OVERFLOW and CFD pain (I’ve been in management & teaching too long …)

– Create demonstration cases for our own Applied Research Group projects

• Demonstrate Large scale big data post processing and effective workflow processes

• Provide a different method to the community

3


AIAA High Lift Prediction Workshop

• Case 1– Coarse, medium, fine and extra fine grid– SARC Turbulence model– Alpha – 7, 16, 18.5, 20, 21, 22.4– Steady– Blind Run up to 10000+ Steps– High Reynolds

• Case 2– High Reynolds Number

• Alpha – 0, 7, 12, 16, 18.5, 20, 21,22.4• SARC Turbulence Model• Fully Turbulent• Blind Run Up to ~10k Steps

– Low Reynolds Number• SA RC Turbulence Model • Fully Turbulent (Blind Run Up to ~10k Steps)• Transition (Menter-Langtry)• k-W-SST • k-W-SST Unsteady

• Case 3– High Reynolds

• Alpha – 0, 7, 12, 16, 18.5, 20, 21,22.4• SARC Turbulence Model• Fully Turbulent• Blind Run Up to ~10k Steps

– Low Reynolds• SARC Turbulence Model • Transition (Menter-Langtry)


OVERFLOW and FV Workflows• Solver - OVERFLOW-2.2e

– Cray XE6– 128-1024 cores (75-85% // Efficiency)– Methodologies

• RHS - Roe upwind • LHS – Scalar Penta• Full Multi-Grid• Full Viscous Terms

• FieldView 13.2 & 14– Cray XE6 – 16 cores– Automated Batch Post-Processing– XDB workflow for surface data

extraction– FVX + gnuplot for 2D plotting– Multi-Window Comparisons


Issue with Large Scale Post-Processing• Writing, reading or copying large

results files takes a long time– Wastes engineering time– Important data is quite small

• Slow network speed– Performance is often unpredictable

• Sharing large results files takes much space and time

• Analyzing many datasets requires repeating many steps

• Graphics hardware outpacingsoftware development

• Data Extracts – XDB Workflows


Volume Data Processing…

READ

Volume Data

CREATE

Post-processing Objects

Compute Objects:- Geometry- Cutting Planes- Iso-surfaces- Streamlines

POSTPROCESS

Actions on Objects:- Visualize/Render- Animate/Sweep Cache- Integrate Forces- Probe/plot values- Output Pics/movies

WRITEXDBFile


XDBFile

…to XDB Workflows

POSTPROCESS

Actions on Objects:- Visualize/Render- Animate/Sweep Cache- Integrate Forces- Probe/plot values- Output Pics/movies

READ/APPEND

XDBFile

CREATE

Post-processing Objects

Compute Objects:- Geometry- Cutting Planes- Iso-surfaces- Streamlines

Full Numerical Fidelity Normals included Smaller files (10X-100X) Lower memory (10X-100X)


HiLiftPW 2 General Workflow• Run the OVERFLOW2 case

128 – 1056 processors • Submit post processing job,

16 procs in batch directly on remote Cray XE6 system– Boundary Surface Extracts– Coordinate cut planes– Velocity Profiles– Surface Streamlines

• Create images, tables and 2D plots– Perform all on the big iron

especially if good graphics connected to system or …

– Move extracts and solver output to local workstation

9


XDB Data reduction

10

Case 1

Grid Size#

Overset Grids

Grid Points(10^6)

Grid File Size

Soln FileSize

Coordinate Cut Plans (3

planes) 7 scalars

Boundary Surface

12 scalar variables

Surface StreamsMedium

seed density, 2 varables

Coarse 44 29.4 758M 1.6G 14.1M 24M 435MMedium 44 69.0 1.8G 3.6G 25.8M 41M 411M

Fine 44 230.7 6.1G 13G 58M 92M 524M

Extra-Fine

44 544.5 15G 29G 98M 160M 530M


CL - CD

11

0

0.5

1

1.5

2

2.5

3

3.5

0 5 10 15 20 25

CL

Alpha (degrees)

Case 1- AoA vs. CL

coarse

medium

fine

exfine

RUN238-Re=15.1e6

0

0.5

1

1.5

2

2.5

3

3.5

0 5 10 15 20 25C

L

Alpha (degrees)

Case 2 - AoA vs. CL

case2a

case2b

case2c

case2a_SST

RUN238-Re=15.1e6

RUN29293-Re=1.35e6

0

0.5

1

1.5

2

2.5

3

3.5

0 5 10 15 20 25

CL

Alpha (degrees)

Case 3 - AoA vs. CL

case3a

case3b

RUN238-Re=15.1e6

RUN29293-Re=1.35e6

0

0.5

1

1.5

2

2.5

3

3.5

0 0.1 0.2 0.3 0.4 0.5 0.6

CL

CD

Case 1-CL vs. CD

coarse

medium

fine

exfine

RUN238-Re=15.1e6

0

0.5

1

1.5

2

2.5

3

3.5

0 0.1 0.2 0.3 0.4 0.5 0.6

CL

CD

Case 2 - CL vs. CD

case2a

case2b

case2c

case2a_SST

RUN238-Re=15.1e6

RUN29293-Re=1.35e6

0

0.5

1

1.5

2

2.5

3

3.5

0 0.1 0.2 0.3 0.4 0.5 0.6

CL

CD

Case 3 - CL vs. CD

case3a

case3b

RUN238-Re=15.1e6

RUN29293-Re=1.35e6


Surface Coefficients• Extract - Boundary Surface

– Slat, Main, Flap, Body– Cp, Cf, Cfx, Cfy, Cfz, Surface

Normals• Further processed on local

workstation• Take slices at the pressure

stations– Use thresholding on the

provided workshop equations and find cell center of resulting polygon

– Output 2D plot data to standard formats

• Automatically create multi plot comparisons using your favorite 2d Plot program (i.e. Gnuplot)

12


Case 1, High Reynolds Number, Medium Grid, a=18.5o


Case 2b, High Reynolds Number, a=18.5o


Case 3b, High Reynolds Number, a=18.5o

-4

-2

0

2 1000 1200 1400 1600

-Cp

x (mm)

PS01

Case 3 High Lift Prediction Workshop Alpha = 18.5 Grid Density = medium Experiment = RUN238_Re=15.1e6

-8

-6

-4

-2

0

2 1200 1400 1600

-Cp

x (mm)

PS02


-10

-8

-6

-4

-2

0

2 1400 1600

-Cp

x (mm)

PS04


-10

-8

-6

-4

-2

0

2 1400 1600 1800

-Cp

x (mm)

PS05


-10

-8

-6

-4

-2

0

2 1600 1800

-Cp

x (mm)

PS06


-10

-8

-6

-4

-2

0

2 1600 1800

-Cp

x (mm)

PS07


-10

-8

-6

-4

-2

0

2 1600 1800

-Cp

x (mm)

PS08


-10

-8

-6

-4

-2

0

2 1800

-Cp

x (mm)

PS09


-10

-8

-6

-4

-2

0

2 1800 2000

-Cp

x (mm)

PS10 Case 3 High Lift Prediction Workshop Alpha = 18.5 Grid Density = medium Experiment = RUN238_Re=15.1e6

-6

-4

-2

0

2 1800 2000

-Cp

x (mm)

PS11 Case 3 High Lift Prediction Workshop Alpha = 18.5 Grid Density = medium Experiment = RUN238_Re=15.1e6


Velocity comparisons• Line Plot Extractions• Directly sample and

export velocity plot data to disk– Experiment vs Case 1

coarse to exfine– Experiment vs Case2

Reynolds number– Experiment vs Case3

Reynolds number– Experiment vs

Case1,2&3 Reynolds number

16

-70-60-50-40-30-20-10

0 10

0 0.5 1 1.5 2 2.5 3

P1 WB L1

-50

-40

-30

-20

-10

0

10

20

0 0.5 1 1.5 2 2.5 3

P1 WB L2

-40-30-20-10

0 10 20 30 40

0 0.5 1 1.5 2 2.5 3

P1 wC L1

-60-50-40-30-20-10

0 10 20

0 0.5 1 1.5 2 2.5 3

P1 wD L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L2

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wD L1

-20

-10

0

10

20

30

40

50

0 0.5 1 1.5 2 2.5 3

P2 wE L1

-40

-30

-20

-10

0

10

20

30

0 0.5 1 1.5 2 2.5 3

P2 wE L2

0

10

20

30

40

50

60

70

0 0.5 1 1.5 2 2.5 3

P3 wE L1

-20

-10

0

10

20

30

40

50

0 0.5 1 1.5 2 2.5 3

P3 wE L2

exp

crse

med

fine

xfine


Experiment vs Case 1

Velocity ProfilesHi Reynolds

7, 18.5 DegreesEffect of Grid Refinement coarse,

medium, fine, extra-fine


Velocity profiles 7 Deg. – Hi Rey

18

-70-60-50-40-30-20-10

0 10

0 0.5 1 1.5 2 2.5 3

P1 WB L1

-50

-40

-30

-20

-10

0

10

20

0 0.5 1 1.5 2 2.5 3

P1 WB L2

-40-30-20-10

0 10 20 30 40

0 0.5 1 1.5 2 2.5 3

P1 wC L1

-60-50-40-30-20-10

0 10 20

0 0.5 1 1.5 2 2.5 3

P1 wD L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L2

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wD L1

-20

-10

0

10

20

30

40

50

0 0.5 1 1.5 2 2.5 3

P2 wE L1

-40

-30

-20

-10

0

10

20

30

0 0.5 1 1.5 2 2.5 3

P2 wE L2

0

10

20

30

40

50

60

70

0 0.5 1 1.5 2 2.5 3

P3 wE L1

-20-10

0 10 20 30 40 50 60

0 0.5 1 1.5 2 2.5 3

P3 wE L2

exp

crse

med

fine

xfine


Velocity profiles 18.5 Deg. – Hi Rey

19

-70-60-50-40-30-20-10

0 10

0 0.5 1 1.5 2 2.5 3

P1 WB L1

-50

-40

-30

-20

-10

0

10

20

0 0.5 1 1.5 2 2.5 3

P1 WB L2

-40-30-20-10

0 10 20 30 40

0 0.5 1 1.5 2 2.5 3

P1 wC L1

-60-50-40-30-20-10

0 10 20

0 0.5 1 1.5 2 2.5 3

P1 wD L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L2

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wD L1

-20

-10

0

10

20

30

40

50

0 0.5 1 1.5 2 2.5 3

P2 wE L1

-40

-30

-20

-10

0

10

20

30

0 0.5 1 1.5 2 2.5 3

P2 wE L2

0

10

20

30

40

50

60

70

0 0.5 1 1.5 2 2.5 3

P3 wE L1

-20

-10

0

10

20

30

40

50

0 0.5 1 1.5 2 2.5 3

P3 wE L2

exp

crse

med

fine

xfine



Effect of Reynolds Number7, 18.5 m/s


Velocity profiles 7 Deg. – Case 2


Velocity profiles 18.5 Deg. – Case 2



Effect of Reynolds Number7, 18.5 m/s


Case 3- 7º Angle of Attack

-70-60-50-40-30-20-10

0 10

0 0.5 1 1.5 2 2.5 3

P1 WB L1

-50

-40

-30

-20

-10

0

10

20

0 0.5 1 1.5 2 2.5 3

P1 WB L2

-40-30-20-10

0 10 20 30 40

0 0.5 1 1.5 2 2.5 3

P1 wC L1

-60-50-40-30-20-10

0 10 20

0 0.5 1 1.5 2 2.5 3

P1 wD L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L2

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wD L1

-20

-10

0

10

20

30

40

50

0 0.5 1 1.5 2 2.5 3

P2 wE L1

-40

-30

-20

-10

0

10

20

30

0 0.5 1 1.5 2 2.5 3

P2 wE L2

0

10

20

30

40

50

60

70

0 0.5 1 1.5 2 2.5 3

P3 wE L1

-20-10

0 10 20 30 40 50 60

0 0.5 1 1.5 2 2.5 3

P3 wE L2

exp

High

Low


Case 3 -18.5º Angle of Attack

-70-60-50-40-30-20-10

0 10

0 0.5 1 1.5 2 2.5 3

P1 WB L1

-50

-40

-30

-20

-10

0

10

20

0 0.5 1 1.5 2 2.5 3

P1 WB L2

-40-30-20-10

0 10 20 30 40

0 0.5 1 1.5 2 2.5 3

P1 wC L1

-60-50-40-30-20-10

0 10 20

0 0.5 1 1.5 2 2.5 3

P1 wD L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L2

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wD L1

-20

-10

0

10

20

30

40

50

0 0.5 1 1.5 2 2.5 3

P2 wE L1

-40

-30

-20

-10

0

10

20

30

0 0.5 1 1.5 2 2.5 3

P2 wE L2

0

10

20

30

40

50

60

70

0 0.5 1 1.5 2 2.5 3

P3 wE L1

-20

-10

0

10

20

30

40

50

0 0.5 1 1.5 2 2.5 3

P3 wE L2

exp

High

Low


Case 1, 2 & 3 18.5º Angle of Attack

-70-60-50-40-30-20-10

0 10

0 0.5 1 1.5 2 2.5 3

P1 WB L1

-50

-40

-30

-20

-10

0

10

20

0 0.5 1 1.5 2 2.5 3

P1 WB L2

-40-30-20-10

0 10 20 30 40

0 0.5 1 1.5 2 2.5 3

P1 wC L1

-60-50-40-30-20-10

0 10 20

0 0.5 1 1.5 2 2.5 3

P1 wD L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L1

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wB L2

-10

0

10

20

30

40

50

60

0 0.5 1 1.5 2 2.5 3

P2 wD L1

-20

-10

0

10

20

30

40

50

0 0.5 1 1.5 2 2.5 3

P2 wE L1

-40

-30

-20

-10

0

10

20

30

0 0.5 1 1.5 2 2.5 3

P2 wE L2

0

10

20

30

40

50

60

70

0 0.5 1 1.5 2 2.5 3

P3 wE L1

-20

-10

0

10

20

30

40

50

0 0.5 1 1.5 2 2.5 3

P3 wE L2

exp

case1

case2

case3


Surface streamlines• Create surface streamlines using Surface Flow Tool

• Execute script based upon FVX -> .fvp files

• Download the resulting pathline file

• Load in to interactive session for exploration or batch process direct to images

• Use multi-windows to compare directly to experimental images

27


Case 1 Surface Streamlines Showing Effect of Grid Resolution Upon the Side of Body Flow, a=18.5o


Case 1, 2b and 3b Surface Streamlines a=18.5o


Case 1, 2a and 3a Surface Streamlines at the Wing-Body junction, Low Reynolds, a=18.5o


Case 2 and 3 Surface Streamlines at Low and High Reynolds Number, a=18.5o


Case 2 & 3 Surface Streamlines Showing Geometry Effect on Upper Wing Surface Flow, Low Reynolds Number a=7, 18.5, 21.0o


Summary•FV XDB workflow enabled

the large data management for this study

•Grid resolution had little effect

•The more complete the geometry the better

•Flow separation very different from experiment

•Separation patterns affected by:

– Pressure Tubes– Reynolds Number

33


Acknowledgements• The work reported herein was funded by an Intelligent Light

research grant.• The author would like to thank Cray Inc. for provided access to their

corporate Cray XE6 computer • Greg Clifford of Cray Inc. for helping to acquire access. • David Whitaker from Cray Inc. for assistance in porting of

OVERFLOW2 to the XE6 and for streamlining the use of FieldView on their system.

• This work could not have been accomplished without good overset grid systems. Thanks goes to Tony Sclafani of The Boeing Company and the workshop team members for providing the grid and input files for OVERFLOW.

• Finally, much gratitude to Steve M. Legensky, General Manager and Founder at Intelligent Light, for his generous effort in his review and discussions of the work presented in the paper and in this presentation.

copyright (c) 2014 intelligent light all rights reservedauthor epd created date 8/6/2014 1:29:44 pm

Documents