CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Main CFD Forum

Linear-Strength Vortex Unsteady Panel Method (2D)

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Display Modes
Old   April 1, 2012, 19:21
Default Linear-Strength Vortex Unsteady Panel Method (2D)
  #1
New Member
 
Anthony
Join Date: Jul 2011
Posts: 9
Rep Power: 5
Sonik_87 is on a distinguished road
Hi everyone,

I am working on a project for the development of a 2D Unsteady Panel Method for Airfoils. As often already suggested in this forum I have been using the book "Low Speed Aerodynamics" which has helped me a lot to produce the Steady State solver (I am writing the code in Matlab at the moment), which works very well.

However I am having great problems in validating my unsteady code.
I have followed the book, mainly the steps are:

1) compute the steady solution at first time step
2) move the airfoil via a prescribed motion path (in my case pitching A*sin(wt))
3) position a concentrated vortex at a certain distance from the T.E., which has unknown intensity GAMMA_W and which is therefore part of the solution
4) use the Kelvin theorem for completing the coefficients matrix A which will be N+2 times N+2 (N+1 gammas on the nodes of the panels) + 1 (latter) unknown shed concentrated vortex
5) compute the RHS (right-hand-side) term considering free stream velocity and wake induced velocity
6) solve the system

The problem seems that the Cp variations are not big enough. I am using a NACA report on wind tunnel experimental data (pitching motion 5 + 5sin(Wt) motion) to check the code. But also the comparison with results contained in the Katz and Plotkin (plot of CL-ALPHA of a 0012) isn't satisfactory.

I compute the Cp at each collocation point via the formula:

Cp_i = 1 - Vel_i^2

where Vel_i is the velocity at the i-th collocation point obtained from the different contributes: free stream (similar to steady state) + kinematic movement (sinusoidal motion) + induced velocities (by panels and by wake).
I consider the free stream V to be equal to 1.Honestly I am not quite sure that all of this is right (talking about Cp calculation).

Moreover I have great doubt about the Kutta condition: in the steady state case the last line of matrix A is set equal to: [1 0 0 0 . . . 0 1] so that vorticity at T.E. is 0:

gamma_1 + gamma_(N+1) = 0

I did not change this in the unsteady code only because I couldn't find any indication on how to modify it (the book says that for small oscillations the steady case condition should still work fine).

I would be really thankful if any of you could help me out, as it is more than a month that I am trying to solve the problem. I have also read many articles like Hancock and Mook's one, or similar, but they don't really say explicitly how to do things.

Thank you very much.

PS: do ask questions if you don't understand some of the procedures I have used, I will be happy to explain more thoroughly.
Sonik_87 is offline   Reply With Quote

Old   April 3, 2012, 13:44
Default
  #2
New Member
 
james nathman
Join Date: May 2011
Posts: 18
Rep Power: 5
blackjack is on a distinguished road
The unsteady Bernoulli eqn applies:

http://web.mit.edu/2.016/www/handout...ion_050921.pdf

(I'm sure there are better descriptions, but this is the first acceptable one I found).

Cp=1-V^2/2+d(phi)/dt

I strongly recommend that you try a wakeless problem such as a cylinder or ellipse. Oscillating in heave is an easier problem than oscillating in pitch.
Removing the wake eliminates the Kutta condition and convection as sources of error.

You should be able to reproduce the apparent mass calculations of Max Munk (See naca reports server for added/apparent mass).

Then use the unsteady Bernoulli equation to determine the time rate of vorticity shedding into the wake. Run a very low frequency - you should get steady-state results.

Describe the airfoil problem in detail: airfoil, frequency, amplitude, pitch axis.
blackjack is offline   Reply With Quote

Old   April 3, 2012, 15:41
Default
  #3
New Member
 
Anthony
Join Date: Jul 2011
Posts: 9
Rep Power: 5
Sonik_87 is on a distinguished road
The problem is I need it for airfoils. At the moment I am using a NACA 0012 and the experimental data with which I need to validate the code is this sinusoidal pitch motion:

5 + 5*sin(wt)

With k=0.2.

I'm not quite sure how to get d(phi)/dt out of my code, because I never handle the potential directly.
Sonik_87 is offline   Reply With Quote

Old   April 4, 2012, 13:52
Default
  #4
New Member
 
james nathman
Join Date: May 2011
Posts: 18
Rep Power: 5
blackjack is on a distinguished road
Sorry for the typo: Cp=1-V^2+2d(phi)/dt

I use the method of Luigi Morino, it applies to Oscillatory or unsteady or steady subsonic conditions. The time derivative is more significant as the frequency (which you haven't specified) increases.

If you have the vorticity on the airfoil surface, check me but isn't

phi=phi_at_leading_edge + integral[ gamma * ds]

The time derivative of phi_at_leading_edge will not contribute to the forces because it's integrated over a closed contour. So you can calculate forces and moments (but you haven't specified what output you are interested in, nor have you stated what cases, if any, you program does accurately calculate)

If you're not getting good results for an airfoil with wake, doesn't it make sense to simplify the problem to one which makes the error obvious? A high-frequency (k=1) wakeless case can determine if the boundary condition and Cp are correct. If you can't calculate a low-frequency (k=0.001) pitch case correctly, it's unlikely higher frequencies will work.
blackjack is offline   Reply With Quote

Old   April 4, 2012, 14:15
Default
  #5
New Member
 
james nathman
Join Date: May 2011
Posts: 18
Rep Power: 5
blackjack is on a distinguished road
Your Kutta condition of equal upper and lower vorticity will not work for an oscillating airfoil which must be allowed to shed vorticity into the wake.

d(wake_circulation)/dt=-Vte*d(wake_circulation)/ds

where Vte is the mean velocity at the trailing edge.
blackjack is offline   Reply With Quote

Old   April 4, 2012, 16:17
Default
  #6
New Member
 
Anthony
Join Date: Jul 2011
Posts: 9
Rep Power: 5
Sonik_87 is on a distinguished road
In an article I have found a quite precise description of the Unsteady Kutta Condition for unsteady cases. The article is:

"Interaction of two-dimensional impulsively started airfoils"
(by WU Fu-bing, ZENG Nian-dong, ZHANG Liang ,and WU De-ming)

In this paper they state: (gamma_1)_k + (gamma_N+1)_k = (gamma_w)_k

(where gamma_w is the value on the extra T.E. panel of constant vorticity, and k indicates the time step)

They also use an iterative method to get delta_k (panel length) and theta_k (inclination of the panel) of the shed vorticity panel. This is then transformed into a concentrated vortex and builds up the wake.

I haven't implemented this yet but I have changed my code setting A matrix in such a way that at the line which expresses the Kutta condition I have:

[1 0 0 . . . 1 -1] which should be exactly what they suggest in the article.

Also I have found an error (which I have now corrected): i.e. I hadn't multiplied the gammas on the panel by the relative panel length to make it consisten with the wake vortices which are obviously concentrated vortices, not linear vortex distributions (this is especially important for the Kelvin condition)!

Things now are quite differente and Cp variations seem to be more realistic: for the mentioned case they go from 0 to approx -6.

Answering to a few questions posed by blackjack:

- I am interested in the Cp, above all;
- I still calculate it via the tangential component of the velocity at each collocation point, without considering the potential's derivative, also because I still don't understand how to (from the gamma or from other quantities) obtain it... should I integrate the gammas on the airfoil to obtain PHI, and then differentiate (= dPHI/dt == (PHI_k-PHI_k-1)/(t_k-t_k-1)) to get the second term of the RHS in "Cp=1-V^2+2d(phi)/dt"... or am I wrong?
- I have this file http://http://www.ar87.com/uni/19810...1981008464.pdf

and at page 7 you can see the data I am trying to get out of my code. So reduced frequency k is 0.2.

What I now get as variations of my Cp is this:
http://www.ar87.com/uni/dCp.jpg
which corresponds in the time line to these points:
http://www.ar87.com/uni/angle.jpg

(I just plotted those 4 points).
Sonik_87 is offline   Reply With Quote

Old   April 5, 2012, 14:04
Default
  #7
New Member
 
james nathman
Join Date: May 2011
Posts: 18
Rep Power: 5
blackjack is on a distinguished road
The link to the .pdf file does not work.

The scheme to find dphi/dt should work.

I have run your case. The time derivative at the leading edge is quite small - it contributes less than 0.003 to Cp but at the trailing edge contributes up to 0.09. This is the same magnitude as the boundary layer you neglect.

The most negative pressure on the airfoil I get is -4.85 without boundary layer. Less with the boundary layer.
blackjack is offline   Reply With Quote

Old   April 6, 2012, 12:34
Default
  #8
New Member
 
Anthony
Join Date: Jul 2011
Posts: 9
Rep Power: 5
Sonik_87 is on a distinguished road
Ok, the test case (experimental data) is this one (the link which wasn't working before):

http://www.ar87.com/uni/testCase.jpg


I have modified my code by adding a function which moves the vortices in the wake under the influence of the induced effect (panel induced velocity and wake self-induced velocity), and the same case produces this:

http://www.ar87.com/uni/dCp1WRUit.jpg

this is again the plot of 4 points on the sinusoidal motion: 0, pi/2, pi, 3/2pi.

I didn't manage to implement the dPHI/dt... how exactly should I calculate it?
What I did was something like this:

if j==1
PHI1(1) = PLEN(1)*(GAMMA(1,i)+GAMMA(2,i))/2;
else
PHI1(j) = PLEN(j)*(GAMMA(j,i)+GAMMA(j+1,i))/2 + sum(PHI1(1:j-1));
end

%dPHI/dt
dPHI(j) = PHI1(j) - PHI0(j);

PHI0(j) = PHI1(j);


where PLEN(j) is the panel length of the j-th panel, GAMMA contains the gamma distribution value at nodes of the panel, PHI should be the value of the potential I want to calculate. Could you help me out in getting it right?

Moreover, do you thing that it makes a great difference if I just use tangential component of the total velocity at each collocation point and calculate the Cp, rather than total velocity + dPHI/dt? (Because in the formula Cp=1-V^2+2d(phi)/dt, V is the module of the velocity, not the tangential velocity. Right?)
Sonik_87 is offline   Reply With Quote

Old   April 6, 2012, 14:45
Default
  #9
New Member
 
james nathman
Join Date: May 2011
Posts: 18
Rep Power: 5
blackjack is on a distinguished road
I don't know the pitch axis for your case: c/4, c/2? Moment is about c/4, c/2?

Your approach to dphi/dt sounds fine. Attached is the potential calculated for a 0012 at steady 10 degrees alpha. This should be useful for verifying that you have calculated the correct potential. The time derivative is then obtained from finite difference of two solutions.
blackjack is offline   Reply With Quote

Old   April 6, 2012, 14:49
Default
  #10
New Member
 
james nathman
Join Date: May 2011
Posts: 18
Rep Power: 5
blackjack is on a distinguished road
Attachment here
Attached Files
File Type: txt phi10.txt (4.3 KB, 11 views)
blackjack is offline   Reply With Quote

Old   April 6, 2012, 17:26
Default
  #11
New Member
 
Anthony
Join Date: Jul 2011
Posts: 9
Rep Power: 5
Sonik_87 is on a distinguished road
Pitch axis is c/4, yes.
Sonik_87 is offline   Reply With Quote

Old   April 7, 2012, 05:41
Default
  #12
New Member
 
Anthony
Join Date: Jul 2011
Posts: 9
Rep Power: 5
Sonik_87 is on a distinguished road
I'm afraid I'm not getting round to calculating the potential.
Thank you for the file, very kind of you. But still I don't get the correct values. I have calculated the 10 steady case for NACA 0012 (which is surely correct as I have correctly validate my steady state code months ago), however I don't know how to get the PHI out of my gammas.

Consider I get the gammas at nodes of the 100 panels on the surface. I have tried different calculations, but none give your results, which I presume are correct.

In Katz&Plotkin they suggest this formula for the potential (in the case of linear vortex distribution):

PHI = +/- gamma1/16*(x1^2 + 2*x1*x2 -3*x2^2)

where gamma1 should be the angular coefficient of the linear formula for the gamma:

gamma(x) = gamma0 + gamma1*(x-x1)


I also tried the formula I have written before, but that produces enormous values (10^24).
So at the moment I am a little bit confused.
Sonik_87 is offline   Reply With Quote

Old   April 11, 2012, 11:00
Default
  #13
New Member
 
Anthony
Join Date: Jul 2011
Posts: 9
Rep Power: 5
Sonik_87 is on a distinguished road
My results seem to be quite correct, but not exactly like in the paper, and the minimum Cp I get is not -4.85 but a bit less.
Sonik_87 is offline   Reply With Quote

Old   April 13, 2012, 05:46
Default
  #14
New Member
 
Anthony
Join Date: Jul 2011
Posts: 9
Rep Power: 5
Sonik_87 is on a distinguished road
I could send my code to anyone of you and you could give it a check... just for some suggestions... Thanks

Last edited by Sonik_87; April 13, 2012 at 06:13.
Sonik_87 is offline   Reply With Quote

Old   June 30, 2012, 17:56
Default Code Request
  #15
New Member
 
Mido
Join Date: Mar 2011
Posts: 16
Rep Power: 5
nw_ds is on a distinguished road
I running through some problems with my code
if Possible could you send me your code
my email : night_wing7@yahoo.com
nw_ds is offline   Reply With Quote

Old   February 4, 2013, 13:48
Default potential calculation
  #16
New Member
 
Pappu
Join Date: Feb 2013
Posts: 1
Rep Power: 0
pappu is on a distinguished road
I an dealing with linear strength vortex panel method.Can anyone of you can help in finding the potential for dphi/dt? If I am doing my programing in local coordinate system(frame fixed in wing as described in KATZ and PLOTKIN) can any body could suggest what should be the V for each panel for the formula Cp=1-V^2/2+d(phi)/dt that is I want to know V is the addition of which terms or simply the V as in steady state case as for i=1:npanel vel=0; for j=1:npanel+1 vel=vel+at(i,j)*gamma(j); end vel=vel+U*tang(i,1)+W*tang(i,2); Cp(i)=1-(vel/Q)^2; end where U=freestream X component velocity and W=freestream Y component velocity
pappu is offline   Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
solution diverges when linear upwind interpolation scheme is used subash OpenFOAM 0 May 29, 2010 01:23
Panel Method Doublet Derivatives Gerrit Groot Main CFD Forum 17 September 17, 2007 09:13
Why use a panel method? gocarts Main CFD Forum 2 April 22, 2007 22:50
Problem building 3D Panel Method Daniel Main CFD Forum 0 September 15, 2006 07:51
calculating missle configuration by panel method Daniel Main CFD Forum 5 January 6, 2003 08:09


All times are GMT -4. The time now is 11:12.