Hello Gurus,

I am trying to implement a CFD simulation on an unstructured mesh (FVM) that requires the explicit calculation of a second-derivative to be used in a source term. This is not the second order diffusion term, but I actually need to calculate a number at each cell. The straightforward approach is to use Gauss divergence theorem and some type of technique for interpolating to the cell boundaries, and computing the first derivative for the cell. Then to repeat the process and calculate the second derivative. But this ends up being far too noisy.

Does anyone have any experience with this, or know of some literature about it? Any help is greatly appreciated.

Keith

You may use the DSC method to get the highly precise value. (see J.Phys, A 33,2000)

Hi, Zhou,

It seems very difficult to find your mentioned paper on DSC method by following your supported information. Could you please give more detailed information (e.g. page number) on this paper?

Thanks

Keith,

I had the same problem with second order derivatives (I needed the velocity for implementing an oldroyd derivative) but I used structured FVM. Applying Gauss two times worked for me, no noise disturbation...

sincerely,

Thomas

Hi,

I have compared two methods to evaluate the second derivative, one is the Green-Gauss type approach, another is by T. Barth(in an AIAA paper, around 1994). My conclusion is that the later one results in better solution on highly streched grids. It appreas that there is more diffusion like effects in the Green-Gauss approach.

Guoping