Embedded Files
sslvpn.asu.edu ---this is the address from lovely Deyzi!!!!
My recent results:
On my recent results, I did a grid refinement case study.
I looked at the cl as I increased the mesh. As the paper suggests that at 80k elements transitioning to 90k elements
the coefficient of lift does not change much. The papers figure 1. shows a grid refinement case study at the stall angle (approximately
16 degrees seems to be the stall angle looking at there cl vs aoa).I am not sure what turbulence model they used to
conduct there grid refinement case study.
On the other hand, I am getting a big difference when I reach 400k elements in the mesh at 5 degree angle of attack.
I chose this angle of attack to do a grid refinement case study due to matching the papers results at low angles
of attack.
Increasing the mesh:
I have attached pictures of the mesh models that were used to plot the curve for the grid refinement case study.
Papers results:
The paper predicts cl is about 0.5 at 5 degrees angle of attack for the naca0012 airfoil. I am predicting to get that value at 400k elements. The paper seemed to get that value at 80k elements. There mesh is an optimized mesh that I am struggling to get.
I am still achieving there predicted cl value though with my mesh at 400k.
My results are the same as the paper at low angles of attack, however at the higher angles of attack the start to not agree
with the experimental data of the naca0012 airfoil.
From my recent results I plan to do the following:
1. Run the next turbulent model, (k-w SST) with the 400k elements to see if I can match the papers predicted results.
The paper suggests that the k-w SST two -equation model had good agreement with the published experimental data.
I intend to see what happens when I begin to use that model.
2. Continue to refine the mesh around the trail of the airfoil to get an optimized mesh as the paper shows.
3. Develop a finer mesh to see if the cl value increases more for even more elements in the mesh.
My pressure and velocity match the paper contours at 3 degrees angle of attack (Contour plots attached).
The Re number they are running their simulations at is 3e6 where there velocity magnitude is 43 m/s.
The paper has velocity magnitude of 62 m/s 1180 Pa at 3 degrees AOA which is what I am getting also.
I changed my journal file to simulate 35 angles of attack and multiplied the components 43*cos(angle) and
43*sin(angle) to match their Reynolds number- I was not doing this before.FLUENT solves for pressure by static plus the atmospheric pressure so I have it set for 0 bar for the static pressure and 1 bar (100,000 Pa) for atmospheric pressure. I am using the pressure-velocity coupled method.
From my recent results I plan to:
1. Run the simulation with more iterations for the higher angles of attack for better prediction to the experimental data at higher
angels of attack. From the paper they state that at higher angles accurate results in commercial CFD codes are not well defined.
2. Begin to validate the k-w and Spalart -Allmaras turbulence model-
the paper suggests that the k-w SST two -equation model had good agreement with the published experimental data.
I intend to see what happens when I begin to use that model.
Nasa Spalart Turbulence Model:
http://turbmodels.larc.nasa.gov/spalart.html
Spalart–Allmaras turbulence model
http://en.wikipedia.org/wiki/Spalart%E2%80%93Allmaras_turbulence_model
The Spalart–Allmaras model is a one equation model for turbulent viscosity. It solves a transport equation for a viscosity-like variable
. This may be referred to as the Spalart–Allmaras variable.
Prandtl number
http://en.wikipedia.org/wiki/Prandtl_number
I got divergence detected in AMG solver so I reduced the relaxation factor from 0.25 to 0.1
and the error went away for the pseduo transient case.
make sure your mesh equivolume skew is less than 0.93-0.95.
Especially important is that you should have low skew in regions where gradients are large.
It is in these regions that errors multiply.
http://en.wikipedia.org/wiki/Lift_(force)
NASA Turbulence Modeling:
http://turbmodels.larc.nasa.gov/naca0012_val.html
NACA 0012 verification and validation case study
https://confluence.cornell.edu/display/SIMULATION/Flow+over+an+Airfoil+-+Mesh
naca 0012 thickness is 12% of its chord.
0.12 * chord, chord = 1m
therefore ==> thickness = 0.12 m
chord = 1m
depth = 1m
Planform area = max thickness * depth
Planform area = 0.12m * 1 m depth
Planform area = 0.12m ^2 (This is taking area of a box)
depth for 2d airfoil?
chord = 1 m
Planform area = chord * thickness
chord = 1 m
thickness = 0.1 m
Planform area = 0.1 m^2
https://confluence.cornell.edu/display/SIMULATION/ANSYS+WB+-+Airfoil+-+All+Pages
steps to setup the 500K mesh CFD simulation on ASU Saguaro SuperComputer
step 1. donload winscp for moving files from your desktop to supercoputer:
http://winscp.net/eng/index.php
step 2.download nomachine
https://www.nomachine.com/AR11B00072
step 2.1. Create an account:
http://a2c2.asu.edu/get-an-account/
1. Download Putty:
http://www.chiark.greenend.org.uk/~sgtatham/putty/download.html
or Download SSH client for windows
file:///C:/Documents%20and%20Settings/mjthomp3/My%20Documents/Downloads/SSHSecureShellClient-3.2.9(2).exe
2. Download WNSCP
3. Download exceed or Xming (they are x forwarding display environments)
from myasu
http://www.straightrunning.com/XmingNotes/
step 3.ssh login with X-forwarding:
ssh -X mjthomp3@saguaro.fulton.asu.edu
qsub -X -I -lnodes=1 -lwalltime=00:15:00
use ansysfluent
use ansysworkbench
fluent
step 4.Read the case file
step 5.Read the jounral file
step 6.Lift and drag txt file and load that into matlab
Matlab:Once you obtain the data: Extrac the 600th iteration to get the value in matlab.
%%% Load the CFD drag data
%
CFDdrag6Million=load('C:\Users\Michael Thompson\Documents\lift\drag.txt')
size(CFD_drag_6Million)
%
% Sort the data
% steady state, 600 iteration is cl value
%
drag_sort=CFDdrag6Million(600:600:21600,:)
cd_value=drag_sort(:,2)
%
%%% Load the CFD lift data
%
CFDlift6Million=load('C:\Users\Michael Thompson\Documents\lift\coeff_lift_history.txt')
size(CFDlift6Million)
%
% Sort the data
% steady state, 600 iteration is cd value
%
lift_sort=CFDlift6Million(600:600:21600,:)
cl_value=lift_sort(:,2)
%
% Plot lift and drag vs AOA
%
AOA=-10:25
size(AOA)
plot(AOA,cd_value,'*r',AOA,cl_value,'b')
legend('drag, 6M mesh size','lift, 6M mesh size')
ylabel('Lift and drag')
xlabel('AOA [deg]')
title('CFD data for 6M mesh')
please do steps 1 and 2 before we meet again.....anything further steps please
Saguaro No Machine for supercomputing:
https://www.nomachine.com/
The aim is to duplicate results for different airfoils
because this helps in the desing of MAV wings
6 Million mesh
%
%%% Load the CFD drag data
%
CFDdrag6Million=load('C:\Users\Michael Thompson\Documents\lift\drag.txt')
size(CFD_drag_6Million)
%
% Sort the data
% steady state, 600 iteration is cl value
%
drag_sort=CFDdrag6Million(600:600:21600,:)
cd_value=drag_sort(:,2)
%
%%% Load the CFD lift data
%
CFDlift6Million=load('C:\Users\Michael Thompson\Documents\lift\coeff_lift_history.txt')
size(CFDlift6Million)
%
% Sort the data
% steady state, 600 iteration is cd value
%
lift_sort=CFDlift6Million(600:600:21600,:)
cl_value=lift_sort(:,2)
%
% Plot lift and drag vs AOA
%
AOA=-10:25
size(AOA)
plot(AOA,cd_value,'*r',AOA,cl_value,'b')
legend('drag, 6M mesh size','lift, 6M mesh size')
ylabel('Lift and drag')
xlabel('AOA [deg]')
title('CFD data for 6M mesh')
FLUENT Theory guide;
http://cdlab2.fluid.tuwien.ac.at/LEHRE/TURB/Fluent.Inc/v140/flu_th.pdf
paper1
CFD SIMULATION OF FLOW PAST MAV WINGS
http://nal-ir.nal.res.in/10242/1/071_G04.pdf
paper2
http://enstrophy.mae.ufl.edu/~mohseni/publications/MyPapers/AIAA2004ChicagoSeptB1.pdf
http://fun3d.larc.nasa.gov/papers/Dissertation_Martin.Jones_10-14-13_final.pdf
http://ho.seas.ucla.edu/wp-content/uploads/2011/04/237.pdf
paper 1:
http://www.isca.in/rjrs/archive/v2/i6/3.ISCA-RJRS-2013-017.pdf
http://www.beta-cae.gr/events/c5pdf/8B_1_kosik.pdf
Page updated
Google Sites
Report abuse