1 Workshop has been scheduled for Jan 2 (3pm-6 pm) Jan 3 (8am-6pm) Shown in the current Aerospace America.

1

Workshop has beenscheduled forJan 2 (3pm-6 pm)Jan 3 (8am-6pm)

Shown in the currentAerospace America

2

Telecon agenda, October 1, 2015

• Review Sept telecon notes• Action items & issues• Administrivia

– Dates– Website– Other happenings

• Results discussion: – Case 2: Euler Results: Andrea Mannarino, Politecnico di Milano– Case 1 FRFs– Case 3: Marcello Righi– Added new results to Case 1 Integrated coefficients plots (included in this file)

• Processing discussion:– Process overview and details– Marcello Righi example FRF processing using matlab scripts?– Feedback from using FRF processing scripts?– Feedback from using damping (logdec) processing scripts?

• Next telecon November 5, 11 a.m.

September telecon summary

• Held on Sept 3, 2015 11 a.m.• Next telecon November, 5 11 a.m. East Coast time in U.S.• Administrative matters

– Telecon slides now uploaded to website– Deadline established for declaring participation as an analysis team: Oct 1– Advertisement blurb submitted to AIAA

• Analysis results– Embraer discussed results for Case 1, both steady and forced oscillation

FRFs– Comparisons of steady coefficients for Case 1 were presented and

discussed. Results separated by turbulence model, grid size and grid source are included in the Sept telecon slides.

• Consistent turbulence model study added to the analysis matrix: Case 2: Flutter at Mach 0.74, 0° angle of attack. For those running RANS analysis, utilize your code’s standard Spalart-Allmaras turbulence model

3

4

Jen’s action items from Oct telecon

• Examine the integrated coefficient results for the unforced system response at Mach 0.85. Compare with Marcello’s. Distribute any results to the anlaysis team email list.

• Distribute data processing software to analysis team email list

• Special session at Aviation? Contact Jeremy and Hamid.

• Look into the symposium that Marcello mentioned. Distribute to DDES analysis teams.

5

Action items and issues

• OCTOBER ???– Update/provide additional results for Case 1 (August and Sept items)– Flutter results comparison for Case 2?

• SEPTEMBER – FRFs of Pressure coefficients/Pitch angle comparison for Test Case 1 (Mach

0.7, 3 degs)– Examine results from lift, drag and pitch moment coefficient comparisons for

Test Case 1 that are in these telecon slides.\– Beta versions of Software

• Damping • FRFs

• AUGUST – Lift and pitching moment coefficient comparison for Test Case 1 (Mach 0.7, 3

degs)

Activity 2014 2015 2016Technical Formulation & Preparation

Workshop kick-off, SciTech 2015

Config, grids, etc. available on-line

Analysis team telecons

Perform analyses.

Commitment to contribute analyses

AePW-2

Update / improve CFD results / code(s)

Perform comparisons, Statistical analyses

AIAA conference paper preparation and presentations

Aeroelastic Prediction Workshop 2 Schedule 6

Workshop at SciTech 2016

Data submittal deadline Nov 15 2015

Aviation 2016SciTech 2017

Key Dates: • Computational Team Telecons: 1st Thursday of every calendar month 11 a.m. EST • Deadline for Commitment to contribute analyses: Oct 1, 2015• Computational Results Submitted by Nov 15, 2015• Workshop: SciTech 2016: Jan 2-3, 2016• SciTech Panel Discussion of Workshop: Jan 2016 at SciTech• 2016 AIAA Aviation Conference Abstract Deadline ~ Nov 1, 2016 • 2017 AIAA SciTech Conference Abstract Deadline ~ June 1, 2016

Aeroelastic Prediction Workshop 2 Schedule 7

Key Dates: • Computational Team Telecons: 1st Thursday of every calendar month 11 a.m. EST • Deadline for Commitment to contribute analyses: Oct 1, 2015• Computational Results Submitted by Nov 15, 2015• Workshop: SciTech 2016: Jan 2-3, 2016• SciTech Panel Discussion of Workshop: Jan 2016 at SciTech• 2016 AIAA Aviation Conference Abstract Deadline ~ Nov 1, 2016 • 2017 AIAA SciTech Conference Abstract Deadline ~ June 1, 2016

Data submittal deadline Nov 15 2015 Aviation 2016

Abstract deadline: November 5 2015

8




• Results discussion: – Case 2: Euler Results: Andrea Mannarino, Politecnico di Milano– Case 1 FRFs– Others that I missed?– Added new results to Case 1 Integrated coefficients plots (included in this file)



9







10

Case 1: Mach 0.7, 3Frequency Response Function Comparison Plots

60% span

UpperSurfac

e

11


60% span

UpperSurfac

e

12


60% span

LowerSurfac

e

13


60% span

LowerSurfac

e

14

Phase Unwrapping

15

Closer Look at Upper Surface Shock Region

60% span

UpperSurfac

e

16


60% span

UpperSurfac

e

17

Example time histories from FUN3D, Upper surface

18


19

Closer Look at Upper Surface Shock Region

60% span

UpperSurfac

e

20


21


Forward Section of the Wing

22


Aft Section of the

Wing

23


Forward Section of the Wing

24







25

Pitching moment coefficient reference point is at 30% chord or 4.8 inches from the leading edge!

Progress Check: Aerodynamic Coefficients

CL CD CM_y Mesh ResolutionMesh (Provided / Own) and

Type Turbulence Model1 0.45100 0.02810 -0.06160 Coarse Provided / Unstructured komega-SST2 0.45100 0.02783 -0.06020 Medium Provided / Unstructured komega-SST3 0.45100 0.02783 -0.06020 Fine Provided / Unstructured komega-SST

4 0.44300 0.02785 -0.05940 Coarse Provided / UnstructuredSpalart-Allmaras One-Equation Model with fv3 Term

(SA-fv3)5 0.45579 0.02846 -0.20021 Coarse Own / Structured Standard Spalart-Allmaras One-Equation Model6 0.45427 0.02813 -0.19926 Medium Own / Structured Standard Spalart-Allmaras One-Equation Model7 0.45310 0.02790 -0.19856 Fine Own / Structured Standard Spalart-Allmaras One-Equation Model8 0.43830 0.02671 -0.14553 Coarse Provided / Unstructured Standard Spalart-Allmaras One-Equation Model9 0.44598 0.02763 -0.14939 Medium Provided / Unstructured Standard Spalart-Allmaras One-Equation Model

10 0.44143 0.02694 -0.14939 Fine Provided / Unstructured Standard Spalart-Allmaras One-Equation Model

11 0.43454 0.02869 -0.18736 Coarse Provided / UnstructuredStandard Spalart-Allmaras One-Equation Model WITH

the ft2 term12 0.44473 0.02889 -0.06080 Fine Provided / Unstructured Standard Spalart-Allmaras One-Equation Model13 0.43536 0.02720 -0.18823 Coarse Provided / Unstructured Standard Spalart-Allmaras One-Equation Model14 0.43711 0.02679 -0.05795 Medium Provided / Unstructured Standard Spalart-Allmaras One-Equation Model15 0.43108 0.02616 -0.18567 Fine Provided / Unstructured Standard Spalart-Allmaras One-Equation Model16 0.40890 0.02570 -0.07180 Coarse Own / Structured Reynolds Stress17 0.40990 0.02560 -0.07190 Coarse Own / Structured SST

18 0.42520 0.02570 -0.07560 Coarse Own / StructuredSpalart-Allmaras One-Equation Model with Edwards

Modification (SA-Edwards)19 0.43068 0.02704 0.09171 Medium Own / Unstructured Standard Spalart-Allmaras One-Equation Model

26

27

28

29

30

31

32

33

34

35







36

Post-processing scripts

• To run the example process:

• launch matlab• load Example_Simplified_Input_File• outline_process_FFT_simplified_input

• You should get the follow plots (and few others)

37

Post-processing scripts

The files gives an example of processing the forced oscillation data, calculating and plotting transfer function estimates (or frequency response function)

The script processes the pressure coefficients for a single span station, separating upper and lower surfaces, assuming that these are input by separate matrices

One set of Fourier analysis parameters are used in this analysis and MUST BE modified by the user to reflect their own data set's requirements

38

Post-processing scripts: Inputs required to be in the workspace prior to running the script

• DescriptionVariable (dimensions)

• Pressure coefficient data at a given span station: cp_mat_upper (# time points x number of upper surface gridpoints)

cp_mat_lower (# of time points x number of lower surface gridpoints)

• Pitch angle time history theta (# of time points x 1)

• Vector of time points, in seconds (physical time, rather than non-dimensionalized or simulation time)

time_vec (# of time points x 1)

• Chord-wise locations of grid points, normalized by the chord (i.e. x/c) xlocs_upper (# of upper surface grid points x 1)

xlocs_lower (# of lower surface grid points x 1)

NOTE THAT THE TIME POINTS FOR THETA AND CP_MATs MUST BE THE SAME

39

Post-processing scripts: User input variable quantities for Fourier processing (edit the script to modify these quantities to be

pertinent to your data set)

str_win string specifying the data windowing to be performed (type help window for list of choices. str_win should be a character string, without a preceding @ sign. e.g. to specify a rectangular window, set str_win = 'rectwin' DEFAULT: str_win='rectwin';

overlap_fac the ratio of number of samples of overlap relative to the periodogram block size (nfft_1). e.g. for 75 overlap, set overlap_fac=0.75;

The other Fourier analysis parameters can be set by specifying any one of the following parameter sets: (Fourier_set = 1, 2, or 3)

SET 1:

nfft_1 Fourier analysis block size or periodogram length (note that this value determines the analysis record length(i.e. total time per analysis) which in term determines the frequency values and spacing) SET 2:

freq_est estimate of the frequency to be isolated.

cycles_per_block number of cycles to include in each analysis block SET 3:

N_seg_0 number of analysis blocks to divide the time history into; i.e. the number of periodograms

40

Post-processing scripts: User input variable quantities for Fourier processing (edit the script to modify these quantities to be

pertinent to your data set)

• Additional input:

freq_lim upper bound on the frequency axis. Matlab runs out of memory on many machines when computing the parameters for large numbers of grid points simultaneously. I have generally been setting this to 100 Hz. The default value, however, is the Nyquist frequency.

41

Results from running outline_process_FFT_simplified_input

For each span

station and each surface,

you get a set of results plots:

Magnitude, Phase, Coherenc

e

42


For each span




e

43


For each span




e

44

Notes on Fourier analysis parameter inter-relationships

T=nfft_1/samp:T(analysis record length in seconds) =nfft_1/samp, where samp is the sample rate in samples/second. samp = 1/ delt, where delt is the spacing of the time vector

delta_frequency (in Hz) = 1 cycle / T seconds = samp / nfft_1 frequency(iii) = (iii-1)*delta_frequencyi.e. freq_axis=(samp/(nfft_1))*[0:(nfft_1-1)];

Note that the Nyquist frequency occurs at the half-waypoint of this freq_axis vector. f_Nyquist = samp/2;

45

Scripts included in the beta release

• read_tecplot_file.m (read point-stacked tecplot file into matlab data structure db_tec. Variables are stored as fields of that data structure)

• outline_process_FFT_simplified_input• periodogram_mimo_jh (multi-input multi-output cross-spectral density

function calculation)• periodogram_jh (for computing 1 cross-spectral density function at a time)• correct_PSD_for_window (called by periodogram program)• overlap_func• overlap_func_approx_nfft• fmult• fdiv• fmash• Example_Simplified_Input_File.mat

Analysis Team Code POCs Email contact

Technion - IIT EZNSS Daniella Raveh [email protected]

FOI, Brno University, Institute of Aerospace Engineering Czech Republic

EDGE Adam Jirasek, Mats Dalenbring,Jan Navratil

[email protected]

NASA FUN3D Pawel Chwalowski, Jennifer Heeg [email protected], [email protected]

NLR NASTRAN Bimo Pranata [email protected]

Indian Institute of Science FLUENT kartik venkatraman [email protected]

ATA Engineering Loci/Chem Eric Blades [email protected]

Embraer S.A. CFD++,CMSoft AERO

Guilherme Ribeiro Begnini [email protected]

Politechnico di Milano Various codes Sergio Ricci [email protected]

Mississippi State MAST Manav Bhatia [email protected]

Zurich University of Applied Sciences (ZHAW, ZUAS)

EDGE, SU2 Marcello Righi [email protected]

General Atomics Aeronautical Systems FLUENT/ANSYS Askar Konkachbaev [email protected]

ANSYS ANSYS Fluent Balasubramanyam Sasanapuri(Krishna Zore, Thorsten Hansen)

[email protected]

University of Strasbourg Yannick Hoarau (Jan Vos) Hoarau [email protected]

University of Michigan SUMad Eirikur Jonsson [email protected]

National Research Council Canada Fereidooni, Amin [email protected]

AePW

-2 A

naly

ses/

Com

mitm

ents

to d

ate

(5/2

9/20

15)

mailto:[email protected]

http://www-imfs.u-strasbg.fr/content/Annuaire?id=49



Analysis Team Code POCs Progress Check / Case 1

Technion - IIT EZNSS Daniella Raveh yes

FOI, Brno University, Institute of Aerospace Engineering Czech Republic

EDGE Adam Jirasek, Mats Dalenbring,Jan Navratil

yes

NASA FUN3D Pawel Chwalowski, Jennifer Heeg yes

NLR NASTRAN Bimo Pranata no

Indian Institute of Science FLUENT kartik venkatraman no

ATA Engineering Loci/Chem Eric Blades no

Embraer S.A. CFD++,CMSoft AERO

Guilherme Ribeiro Begnini yes

Politechnico di Milano Various codes Sergio Ricci no

Mississippi State MAST Manav Bhatia yes

Zurich University of Applied Sciences (ZHAW, ZUAS)

EDGE, SU2 Marcello Righi yes

General Atomics Aeronautical Systems FLUENT/ANSYS Askar Konkachbaev no

ANSYS ANSYS Fluent Balasubramanyam Sasanapuri(Krishna Zore, Thorsten Hansen)

yes

University of Strasbourg Yannick Hoarau (Jan Vos) no

University of Michigan SUMad Eirikur Jonsson yes

National Research Council Canada Fereidooni, Amin no

AePW

-2 A

naly

ses/

Com

mitm

ents

to d

ate

(5/2

9/20

15)




48

Potential special sessions for AePW-2

• International Forum on Aeroelasticity & Structural Dynamics, 2017 Lake Como

• AIAA Aviation Conference: June 2016• AIAA SciTech: Jan 2017• NATO AVT-246 specialists meeting, September 2016

49

• Preliminary Meeting Announcement

• and

• CALL FOR PAPERS

• for the

• AVT-246 Specialists’ Meeting (RSM)

• on

• Progress and Challenges in Validation Testing for Computational Fluid Dynamics• Organized by the Members of the

• APPLIED VEHICLE TECHNOLOGY PANEL (AVT) AVT-246

– to be held in Zaragoza, Spain 26-28 September 2016

• Contributions and participation are invited from NATO Nations, Australia and Sweden only• Final Deadline for submission of abstracts is 1 October 2015

50

Analysis team map

51

Telecon agenda, September 3, 2015

• Review July/August telecon notes• Action items & issues• Administrivia

– Dates– Website– AIAA coordination– Other happenings

• Results discussion: Embraer S. A.

• Next telecon: October 1, 11 am

52

Frequency response function estimates using fourier analysis

• references:• Oppenheim & Shafer, Discrete-time signal

processing• Jay Hardin, NASA RP-1145, Introduction to time

series analysis• Bendat & Pierson, Engineering applications of

correlation and spectral analysis

Frequency response functions (FRFs) magnitude calculation example

Here,x = pitch angley = Cp

• 1 FRF for each pressure transducer or grid point

Frequency, Hz

Mag

nitu

de o

f FRF

, Cp/

q

Pressure / excitation: At frequencies where there is no excitation, the calculation is dividing by 0’ish numbers, making the FRF a large amplitude noisy response

)()(

)( ,, fPSD

fCSDfFRF

x

xyxy



• 1 FRF for each pressure transducer or grid point• Examine values only at the excitation frequency

Frequency, Hz Excitation frequency , in this example, ~80 Hz

Mag

nitu

de o

f FRF

, Cp/

q

)()(

)( ,, fPSD

fCSDfFRF

x

xyxy



• 1 FRF for each pressure transducer or grid point• Examine values only at the excitation frequency• Plot the results for all transducers on a single plot, as a function of chord location

Frequency, Hz Excitation frequency ~ 80 Hz

Mag

nitu

de o

f FRF

, Cp/

q

Mag

nitu

de o

f FRF

, Cp/

q

x/c

Evaluated at the excitation frequency ~ 80 Hz

)()(

)( ,, fPSD

fCSDfFRF

x

xyxy

56

Frequency response function estimates using Fourier analysis

)()(

)(,,)( ,

, fPSDfCSD

fxInputyOutputfFRF

x

xyxy

)(*)())((

)(,,)(

2,

, fPSDfPSDfCSDabs

fxInputyOutputfMSCohere

xy

xyxy

FRFy,x(f): frequency response function estimate for response y due to input xPSDx(f): Power spectral density function estimate for a time history xCSDy,x(f): Cross spectral density function estimate for a response y due to an input xMSCoherey,x(f): or g2

y,x(f): mean squared coherence estimate for a response y due to an input xFFTx(f): discrete Fourier transform of time history x, at frequencies f. Time history x has length nfft samples and a sample rate of samp.f: frequency, Hzk: integer index of frequencysamp: number of samples per second = (1/time step size)nfft: number of samples (points in time) in the data record (time history) being analyzed by the Fourier transform Note that this will be the block size or the periodogram length when the periodogram (block averaging) method is employed

57

Frequency response function estimates using Fourier analysis

sampnfftfFFTfFFT

fCSD xyxy *

)'(*).(*2)(,

sampnfftfFFTfFFTfPSD xx

x *)'(*).(*2)(

sampIblocklengthfFFTfFFTfPSD xxIblock

x *)_()'(*).(*2)(_

58

Discrete Fourier Transform

NFFT

n

NFFTnkj

x enxkFFT1

)1)(1(2*)()(

)1(*)( kNFFTsampkf

]:1[ NFFTk

59

Code snippets [Pxx_1, frsw, junk, junk, junk,junk]=periodogram_jh_v4(thist_in,thist_in,nfft_1,N_overlap_1,str_win,samp,0); [Pyy_1, frsw, junk, junk, junk,junk]=periodogram_jh_v4(thist_out,thist_out,nfft_1,N_overlap_1,str_win,samp,0); [Pxy_1, frsw, junk, junk, junk,junk]=periodogram_jh_v4(thist_in,thist_out,nfft_1,N_overlap_1,str_win,samp,0); Txy_1=Pxy_1 ./ Pxx_1; Cxy_1 = (abs(Pxy_1).^2)./(Pxx_1.*Pyy_1);______________________________________________________________________________________________________________________________________________________________________________________________function [Pxy, freq_axis, fft_x_amplitude, fft_y_amplitude, N_blocks, Save_data]=periodogram_jh_v4(thist_x,thist_y,nfft_1,N_overlap_1,str_win,samp,flag_amplitudes)

freq_axis=(samp/(nfft_1))*[0:(nfft_1-1)];

for iii=1:N_blocks; ibeg=(iii-1)*(nfft_1-N_overlap_1) + 1; iend=ibeg+nfft_1-1; % Detrending added for each segment for experimental data. This is % not in the CFD processing thist_xa=detrend(thist_x(ibeg:iend),'constant'); thist_ya=detrend(thist_y(ibeg:iend),'constant'); thist_xwin=win_1 .* thist_xa; thist_ywin=win_1 .* thist_ya; fftx_1a=fft(thist_xwin); ffty_1a=fft(thist_ywin); %%%Pxy_1a=2*(fftx_1a .* conj(ffty_1a))/ nfft_1 / samp; Pxy_1a=2*(ffty_1a .* conj(fftx_1a))/ nfft_1 / samp; Pxy_sum=Pxy_sum + Pxy_1a;end;Pxy=Pxy_sum / N_blocks;

NOTE: In the coding, the meaning of the x and y variables areREVERSED relative to the derivations on the previous slides!!!!% thist_x is numerator quantity time history (i.e. the sensor in an FRF)% thist_y is the denominator quantity time history (i.e. the actuator in an FRF)

60

Correcting for windowing- Needed if you are comparing PSDs or CSDs instead of FRFs

(JH function correct_PSD_for_window.m)

% Pxx_in Given the PSD computed the usual Jen normalizations:%% PSD_junk=(fft_x .* conj(fft_x)) / (nfft)/samp;% PSD_in= [PSD_junk(1); 2*PSD_junk(2:end-1) ; PSD_junk(end)];%% win1 is the time history (Impulse response function) of the window%% nfft is the time history length, or the Fourier analysis block length%% samp is the sample rate in samples/second = 1/ delta_t

function [PSD_corrected,PSD_scaled,Sine_amp]=correct_PSD_for_window(Pxx_in,win1,nfft,samp);

% PSD_scaled:% Normalizing by NPG_jh will give the same answer for the PSD as returned by the% pwelch function and the cpsd calculation. i.e. the _scaled values of% the PSD are equivalent to those produced by pwelch and cpsd. These% values represent the PSD values normalized by the Noise Power Gain.%% PSD_corrected:% In order to remove the magnitude scaling effect of the window from the% PSD, it is required that the scaled PSD be multipled by the Equivalent% Noise BandWidth.%% or, all of this normalization can be done in 1 step. See the equation% for PSD_manual_Equiv_to_unwindowed_check.% % Sine_amp is the amplitude of the sinusoidal time history, calculated% from the PSD results

61

Correcting for windowing - Needed if you are comparing PSDs or CSDs instead of FRFs

(JH function correct_PSD_for_window.m)

norm_win1=sum (win1 .* win1) ^(1/2);NPG_jh=norm_win1^2/nfft;PSD_scaled = (1/NPG_jh) * Pxx_in; Window_norm=norm(win1); % to normalize such that the PSD produced by a windowed time history is% the same as that produced by the orignal sine wave time history,% multiply by ENBW (equivalent noise band width)ENBW = nfft*sum(win1 .* win1 ) / sum(win1)^2;PSD_corrected = PSD_scaled * ENBW;

function [PSD_corrected,PSD_scaled,Sine_amp]=correct_PSD_for_window(Pxx_in,win1,nfft,samp);

62

Telecon agenda, August 6, 2015

• Review July telecon notes• Action items & issues• Administrivia

– Dates– Website– AIAA coordination– Other happenings

• Results discussion

• Post processing topics– Frequency response function calculations– Damping calculation (loss factor)

• Next telecon Sept 3, 11 a.m.

63

Damping calculation using logarithmic decrement with moving block technique

• This isn’t the only method to determine the modal damping.

• It is based on the assumption of a single degree of freedom with damped harmonic oscillation.

• Damping should ideally be calculated using the pitch angle.

• The method can be used on the generalized displacement results, but note that for cases with any significant amount of dynamic pressure, the modes are coupled. They have to be combined to represent a physical quantity.

64

Logarithmic Decrement applied to AePW-2 analysis results: Outline of method

• Compute pitch angle• Subset the data by user-interactive data points for

beginning and end points• Remove the mean• Determine zero crossings• Interpolate to find improved detrending parameters• Utilize the absolute value of the detrended time history • Take logarithm of peak values for each ½ cycle block• Perform Least squares fit to the data• damping=-1*Linear term coefficient / Mean_freq_rad;

65

66

67

68

69

70

71

General material and prior telecon summaries

72

Updated analysis parameter table

August telecon summary

• Held on August 6, 2015 11 a.m.• Next telecon October 1, 11 a.m. East Coast time in U.S.• Administrative matters


• Analysis results– Jennifer Heeg showed (July telecon) unforced unsteady results for Case 3 (Mach

0.85, 5°); Shock motion of ~9% of the chord for the unforced (no excitation) system is very similar using:

• EZNSS hybrid DDES (based on k-w SST) , shown by Daniella Raveh on June telecon• FUN3D RANS + SA • FUN3D URANS + SA

– Jennifer Heeg discussed (August telecon) frequency response function (FRF)• Consistent turbulence model study added to the analysis matrix: Case 2:

Flutter at Mach 0.74, 0° angle of attack. For those running RANS analysis, utilize your code’s standard Spalart-Allmaras turbulence model

• Discussed the format and content of the panel discussion for SciTech

73

July telecon summary

• Held on July 2, 2015 11 a.m.• Next telecon August 6, 11 a.m. East Coast time in U.S.• Administrative matters


• Analysis results– Krishna Zore from ANSYS shared results- phase difference relative to experimental

data noted; may be definition – Jennifer Heeg showed unforced unsteady results for Case 3 (Mach 0.85, 5°); Shock

motion of ~9% of the chord for the unforced (no excitation) system is very similar using:• EZNSS hybrid DDES (based on k-w SST) , shown by Daniella Raveh on June telecon• FUN3D RANS + SA • FUN3D URANS + SA

– Daniella Raveh from Technion showed results on the June telecon for Case 3: EZNSS RANS solutions showed dependence on the turbulence model

• Consistent turbulence model study added to the analysis matrix: Case 2: Flutter at Mach 0.74, 0° angle of attack. For those running RANS analysis, utilize your code’s standard Spalart-Allmaras turbulence model

• Discussed the format and content of the panel discussion for SciTech74

June telecon summary

• Held on June 11, 2015 11 a.m.• Next telecon July 2, 11 a.m. East Coast time in U.S.• Administrative matters

– Analysis team matrix updates continue– Introduced SciTech panel discussion

• Analysis results

Marcello Righi, Zurich University of Applied Sciences (ZHAW, ZUAS) – Case 1: showed results using Edge and SU2

• Unforced system shown as both average results from dynamic case and steady analysis• Frequency response functions at forcing frequency & at higher harmonics; showed

disagreement in the shock/divot region

Daniella Raveh, Technion: – Cases 2: Varied time step size, temporal convergence criteria and turbulence model

• Flutter frequency was slightly lower with a finer mesh; • temporal convergence study showed increased damping with decreased time step size; “good enough” declared at time step

size of 0.00024 seconds• turbulence model changed the damping• Solution hasn’t converged to an oscillatory behavior at 1.5 seconds (~ 6 cycles); more iterations (global time steps) are needed

– Case 3: hybrid DDES shows unsteady flow with shock motion

75

76

May telecon summary

• Held on May 7, 2015 11 a.m.• Next telecon June 11, 11 a.m. East Coast time in U.S.• Discussed administrative matters

– AIAA coordination: Workshop will be held Saturday Jan 2 (3pm-6pm) & Sunday Jan 3 (8am-6pm)

– Workshop process– Workshop agenda– Discussed having a panel / discussion session at SciTech- during the

conference week– Analysis team matrix updates continue– Suggested face to face at AIAA Aviation conference- not a lot of anticipated

participation• Corrected & updated workshop information from May

– Units on stiffness values• Kh = 2637 lb/ft = 219.75 lb/in = 219.75 slinch/sec^2• Ktheta = 2964 ft-lb/rad = 35568 in-lb/rad= 35568 slinch-in^2/s^2/rad

– Corrected Reynolds number for Case 1 (Mach 0.7, 3°)• Rec = 4.56x106; Re = 3.456x106

77

AIAA Interactions

Approved and signed off byBruce Willis, Chairman of Structural Dynamics Technical CommitteeMegan Scheidt, Managing Director of Products and Programs

78

Envisioned Workshop Process for Analysis Teams (May, 2015)

• Perform analyses• Submit results • Prepare informal presentations for workshop• SciTech 2016

– AePW-2• Present results• Results comparisons• Discussion of results• Path forward

– Panel discussion???• Re-analyze• Publish at special sessions of conferences (which

conferences?)• Publish combined journal articles

79

AePW-2 Agenda Thoughts

• Incorporate fresh perspectives in how we organize the workshop• Following past workshops:

– Introductory material• Welcome & overview• Experimental data set• Geometry & grid system overview

– Participant presentations– Workshop data summary & discussion– Path forward, re-analysis discussions

• Propose a roundtable discussion (1 hour? 2 hours?) for the SciTech conference a few days after the workshop– Brief overview of the activity– Summary of the data comparisons– Panel containing willing and eager analysis team members

80

April telecon summary

• Held on April 2, 2015 11 a.m.• Next telecon May 7, 11 a.m. East Coast time in U.S.• Updated analysis parameters matrix; uploaded to website• Experimental data was added to website• List of analysis teams produced• Discussion of workshop dates• Experimental data reduction showing “divot” in the FRFs to

likely be physical• Pawel showed animation of flutter solution at Mach 0.74

using FUN3D

81

March telecon summary• Website address: http://nescacademy.nasa.gov/workshops/AePW2/public/• Held on March 12, rather than March 5 (with the usual March daylight savings time issues)• Next telecon April 2, 11 a.m. East Coast time in U.S.• SU-2 doesn’t have existing FSI capability.(Melike and Dave Schuster to talk about this?)• Block-structured grids from AePW-1 are available, generated by Thorsten Hansen at ANSYS. (Thorsten

and Pawel will work together to make those available on the new website.)• The molecular weight of R-134a isn’t the same as a standard property table shows (102 g/mol). The

value derived using the listed properties is more like 98 g/mol. This is due to the practical issue of gas purity that is achieved in the wind tunnel. The values on the table are from the test data, where the purity was likely 95%’ish. (Pawel will add a line for molecular weight to the analysis parameters table.)

• Add the following to the table of analyses:– ATA Engineering (Eric Blades will run LoPsiChem)– AFRL (Rick Graves will run FUN3D)– Milano Polytechnico (Sergio Ricci will run numerous codes)

• Please send comments regarding the distributed slides. In particular, are you okay with the abstract submittal form?

• With regard to submitting data to the workshop for comparison:– Can you provide results in matlab?– How do you feel about providing them in a data structure in matlab?

• Doublet lattice aeroelastic solution results:– Bimo and Jen will work to present the results to date at the next telecon– We will put the bulk data file, including the aero model and the flutter cards on the web site. This can serve as a basis for those who might want

to use correction methods, etc.

• Temporal convergence results– Organizations may not have the resources to perform the temporal convergence study for all grids. It is suggested that this be done for a grid

resolution where things look to be spatially converged. Experience at NASA has shown qualitatively different results for the unstructured coarse grid than those observed for the finer grid resolutions.

– The flutter results at low Mach number (Mach 0.74) have shown great variation with regard to time step size. The predicted aeroelasticity stability of the system has been shown to be a function of the time step size and the subiteration convergence level.

82

March telecon summary• Website address: http://nescacademy.nasa.gov/workshops/AePW2/public/• Held on March 12, rather than March 5 (with the usual March daylight savings time issues)• Next telecon April 2, 11 a.m. East Coast time in U.S.• SU-2 doesn’t have existing FSI capability.(Melike and Dave Schuster to talk about this?)• Block-structured grids from AePW-1 are available, generated by Thorsten Hansen at ANSYS. (Thorsten

and Pawel will work together to make those available on the new website.)• The molecular weight of R-134a isn’t the same as a standard property table shows (102 g/mol). The

value derived using the listed properties is more like 98 g/mol. This is due to the practical issue of gas purity that is achieved in the wind tunnel. The values on the table are from the test data, where the purity was likely 95%’ish. (Pawel will add a line for molecular weight to the analysis parameters table.)

• Add the following to the table of analyses:– ATA Engineering (Eric Blades will run LoPsiChem)– AFRL (Rick Graves will run FUN3D)– Milano Polytechnico (Sergio Ricci will run numerous codes)

• Please send comments regarding the distributed slides. In particular, are you okay with the abstract submittal form?

• With regard to submitting data to the workshop for comparison:– Can you provide results in matlab?– How do you feel about providing them in a data structure in matlab?

• Doublet lattice aeroelastic solution results:– Bimo and Jen will work to present the results to date at the next telecon– We will put the bulk data file, including the aero model and the flutter cards on the web site. This can serve as a basis for those who might want

to use correction methods, etc.

• Temporal convergence results– Organizations may not have the resources to perform the temporal convergence study for all grids. It is suggested that this be done for a grid

resolution where things look to be spatially converged. Experience at NASA has shown qualitatively different results for the unstructured coarse grid than those observed for the finer grid resolutions.

– The flutter results at low Mach number (Mach 0.74) have shown great variation with regard to time step size. The predicted aeroelasticity stability of the system has been shown to be a function of the time step size and the subiteration convergence level.

Feb Telecon Notes• Attendees list (to be added)• Suggested adding to website:

– Participating teams and matrix with contact information– Experimental data (Action item taken by Jen.)

• Request made that the frequency response function information be available in both rectangular form (Re and Im components) as well as in polar (Mag and phase) form. (Action item taken by Jen.)

• Experimental results for Case 1. In the FRF magnitude, there is a sawtooth near the leading edge. What is the source of that? Physical? Sensor issue? (Action item taken by Jen.)

• Grids: structured grids were generated by NASA in plot3D format using Pointwise. The gridding guidelines still include the RSW and HIRENASD from AePW-1. Need to revise them so that they are not confusing. Revisit them also with regard to the Reynolds number.

• Nonlinear effects and LCO:– Discussion regarding hysteresis and identification of the neutral stability point– Discussion about experimental data sets, including a DLR study on LCO where there were trends with Mach

number• Process:

– Think about what questions we are trying to answer– How do we tell the organizing committee that we are participating by performing analyses? Is there a website

sign up or abstract submittal form that we mail? • Note: following the end of the telecon, as the webex window was closing… it was noted that there

were some questions and/or comments on the webex communication window. Apologies for not noticing them. The window closed before we could stop it. We are not smart enough to figure out the now-erased questions. Can you ask them again?

• Next telecon March 5, 11 a.m.

84

Mini-abstract from AePW-1

MRL and USF Contribution to AePW - 1N. N. Thusiast_

Multielement Research Lab, Mail Stop 000, Happy Forks, VA 00000 email: [email protected], (777) 777-7777Soar N. Air†

University of Southern Flight, Mail Code 98765, Lofty Heights, TX 00000 email: [email protected], (888) 888-8888

We intend to participate in the AePW-1, to be held April 21-22 2012 in Honolulu, HI. We plan to perform the following sets of computations:

Configuration 1 – RSW , Steady Case, i. M=.825, =2 degCode: RANS-CFD-3DGrid: Str-OnetoOne-C-v1 (supplied by AePW-1 committee)Turbulence model: Menter SST

Configuration 1 – RSW , Unsteady Case, i. M=.825, =2 deg, 10 Hz Same as above

Configuration 2 – BSCW, Steady case, M=.85, =5 deg, 10 Hz Same as above

Configuration 2 – BSCW, Unteady case, M=.85, =5 deg, 20 Hz Same as above

Configuration 3 - HIRENASD Configuration, steady, M=.8, Re=7 million, =1.5 degCode: RANS-CFD-3DAeGrid: Str-OnetoOne-C-v1 (supplied by AePW-1 committee)Turbulence model: S-A

We plan to submit our results electronically by the March 20, 2012 deadline to the AePW-1 committee. RANS-CFD-3DAe is a Reynolds-averaged Navier-Stokes code developed by Et et al.,1 widely used at the

Multielement Research Lab. It is specifically formulated to work on three-element wing configurations. Ituses point-matched grids, and is an upwind finite-volume structured code.LES-CFD-3D is a large-eddy simulation code developed at the University of Southern Flight.2 It employs 6th order central differencing in space and 3rd order temporal differencing, along with 9th order

explicit filtering.

ReferencesEt, H., Cet, P., and Era L., “Description of RANS-CFD-3D,” Journal of Codes, Vol. 6, No. 5, 1994, pp. 5– 21.Author, A. and Author B., “Description of LES-CFD-3D,” Journal of Lengthy Papers, Vol. 9, No. 2, 2008, pp. 22–1021.

_ Corresponding Author. Senior Research Scientist, High Lift Branch.† Professor and Chair, Dept. of Aeronautical Engineering.1 of



1 Workshop has been scheduled for Jan 2 (3pm-6 pm) Jan 3 (8am-6pm) Shown in the current Aerospace America.

Documents