[DAVIDRASC]

Hi guys!
does anyone know how to apply normal boundary condition to an oblique wall in Finite Volume code?
when alpha is zero and beta is 90 we can write PHI(NI,J)=PHI(NI-1,J) within a loop but what should we do about problems with oblique wall?


[Picture is uploaded in this address:]
https://www.cfd-online.com/Forums/attachment.php?attachmentid=74306&stc=1&d=15794540 05

[FMDenaro]

n.Grad phi = known = nx *d phi/dx +ny*d phy/dy



you have a relation involving both derivatives at the boundary

[DAVIDRASC]

Originally Posted by FMDenaro

n.Grad phi = known = nx *d phi/dx +ny*d phy/dy



you have a relation involving both derivatives at the boundary

So thanks Filippo but i think by this relation we would have derivatives in x and y direction that cause some problems when implementation in code. Because the cells and i , j indexes are created in ksi and eta coordinate. Isnt that so?

[FMDenaro]

Originally Posted by DAVIDRASC

So thanks Filippo but i think by this relation we would have derivatives in x and y direction that cause some problems when implementation in code. Because the cells and i , j indexes are created in ksi and eta coordinate. Isnt that so?

I just wrote the normal derivative in cartesian components as example, you can express the relation n.Grad phi = g as better it fits your coordinates. The key is that you have a combination of the derivatives bu means of the components of the unit normal vector to the boundary

[DAVIDRASC]

Originally Posted by FMDenaro

I just wrote the normal derivative in cartesian components as example, you can express the relation n.Grad phi = g as better it fits your coordinates. The key is that you have a combination of the derivatives bu means of the components of the unit normal vector to the boundary

It means that should I use chain derivative rules to get a relation that fits to my Geometrical coordinates?

[LuckyTran]

Originally Posted by DAVIDRASC

It means that should I use chain derivative rules to get a relation that fits to my Geometrical coordinates?

You only need to chain rule when you are switching from your coordinate to another coordinate system. If you're already in ksi and eta then the same formula holds. You're taking the product of the normal vector and the gradient (n.grad()), no matter the expression in any coordinate system.

[DAVIDRASC]

so thanks. I did it!
https://www.cfd-online.com/Forums/at...1&d=1579547532

Thanks for your guidance.