Numerical Diffusion in SIMPLER.
 

Hi,

I am working on a problem of analysing convection in vertical co-axial cylinders. The innner cylinder is not completely "immersed" in the outer one. The outer cylinder has a prescribed iso-thermal wall temprature, and the two cylinders are exchanging engergy via radiation and convection.

I used a SIMPLER code to solve the flow field. I also have an experimental setup for this config. The exp. conf shows some convection cells that the numerical code does not predict. Also, the agreement between the tempratures is not very good, specially at the top of the innner cylinder.

I read that SIMPLER has inherent numerical diffusion that may wipe out convection cells. How do I overcome this problem? Can I try to change the differencing scheme?

Thanks in advance, Regards, Ammar.

Re: Numerical Diffusion in SIMPLER.
 

(1). What is your mesh size and the distribution of the cells? (2). Is the flow laminar or turbulent? (3). You need to increase the total number of cells (mesh size) until the solution is mesh independent. (4).If the flow is turbulent, you need to try the low Reynolds number model. (5). The numerical solutions should be different from the test results. Don't try to match the numerical solutions with the test data, unless your numerical solution is mesh independent.

Re: Numerical Diffusion in SIMPLER.
 

There is nothing in SIMPLER what would foorce it to have inherent numerical diffusion. If the details in your solution do not match the experiments, then either:

1. Your mesh is not fine enough (J.Chien's post).

2. Your convection differencing scheme is too diffusive. Try some higher order schemes (e.g. the second order upwind or better QUICK if you use a structured mesh).

3. There is something wrong with your model. At first, I would try to check if your model conserves mass and energy.

regards

DML

Re: Numerical Diffusion in SIMPLER.
 

Hi, Ammar Hakim

If you use the Patankar's Power Law Scheme (as it is in classical Patankar's SIMPLER) then you have to know that the Power Law cann't find small vortices (so you have to use very big grid or other scheme, for example QUICK)

Best regards,

Evgueny