CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   A problem of panel mathod code. (

lily June 2, 2001 04:37

A problem of panel mathod code.
I am developing a fortran program to sovle the unsteady 2-D potential flow problem with the classical panel mathod.Combined constant strength doublet and source was chosen to distribute along the airfoil,and the wake was modeled by constant strength doublet.But now ,the lift coefficient vary with diffrent airfoils.who can help me?

elson June 2, 2001 09:28

Re: A problem of panel mathod code.
hello Lily. I use the source panel method to simulate the airfoil. Then I would like to know, if you put the Kutta Condition in the Trailing edge.If you do not put, maybe the problem is this. Elson

Lily June 2, 2001 11:54

Re: A problem of panel mathod code.

Thanks a lot for your replying. But I still have the problem.Would you please help me more?

1).My program was developed based on the theory of the following book: "Low-Speed Aerodynamics, From Wing Theory to Panel Methods" By Katz. J. & Plotkin Allen.

2).I am sure I had put the Kutta condition in the trailing edge.

3).In my program,the source/doublet panel method was used to simulate the 2-D unsteady flow.And it is based on the Dirichlet boundary condition.

4).Problem: When a elliptic airfoil was chosen,doing heaving motion with small amplitude,the lift coefficient decreased while the panel number increased.Isn't it unreasonable?

If you'd like to help me,I can send you my code so you can have a look at it and check it for me.Or if you have some panel method code,would you please send me a copy?

Best wishes for u!


Trac June 3, 2001 20:03

Re: A problem of panel mathod code.
Hi Lily,

How many elements are you using on the airfoil? Is it possible you are just increasing the accuracy by increasing the number of panels?


Lily June 3, 2001 22:35

Re: A problem of panel mathod code.

Thanks for your replying.

When I use 36 panels on the airfoil,and dt=0.01,the cl-t line was close to the theoretical curve. But when I use 72 panels and 120 panels,the lift coeffecient was smaller than the theoretical value.That is to say,when panel numbers increased,the accuracy decreased.

Today,I use a NACA0012 airfoil instead of the elliptic airfoil,and got a better result.But,the difference between different panel numbers still exists.

Could anyone give me some more suggestion?

Trac June 3, 2001 22:40

Re: A problem of panel mathod code.
Hi Lily,

You might want to try a few different panel numbers and plot the Cl vs panel number. It might be converging towards a value as you increase the number of panels, and that first one might just be a coincidence. What angle of attack are you using?


Lily June 4, 2001 02:22

Re: A problem of panel mathod code.

I am just doing a validation of small amplitude heaving motion.So I can't plot the cl vs panel numbers. I only can plot the cl vs time under specific panel numbers.And I can also plot the theoretical cl vs time under the same conditions.Some parameters of the motion is: k=1.0, heaving amplitude=0.1c, angle of attack=0.

Now,the problem is: more panels I use,the cl depart more from the theoretic values and get smaller.Can you tell me what can I do?


Chetan June 21, 2001 01:01

Re: A problem of panel mathod code.
When I studied the panel method, I learned that this was an inviscid method. The fact that your panel method may not consider viscous effects could be one reason why you should expect some differences between emperical results. But you are comparing your results to theoretical results... what are these theoretical results? Are they hand calcs or DNS 3d turbulence modelling on a supercomputer? Results of a panel method would fall somewhere in between as far as accuracy is concerned. Also note that a 2d panel method will not take into account 3d effects including tip vortices. I agree that you should compare your results with varying number of panels first and then worry about differences between your solution and your so called theoretical solution. For a given period of time, average out the coefficient of lift. Now graph the time averaged results as a function of number of cells along an x axis and number of cells along the y axis. You should then find be able to see the type of cell spacing needed to get a converged solution. If you want, you can send me your code. I also have written a panel method program (2d-steady) that you could look at if you like.

All times are GMT -4. The time now is 15:00.