Radiation in STAR
 

Dear All,

I have been a CFD analyst, at least part-time, for several years now, mainly on Fluent, the robustness of this code is a constant irritation to me.

I am now starting a new job, in which I will have a say in the decision which code to get, and wondered if anyone could give me some upfront advice.

From what I heard, I am definitely interested in trying STAR, but the modelling will mainly involve thermal stuff, and radiation is important. Fluent has a smart radiation model which I used before, the "Discrete Ordinates" model. This dispenses with the additional overhead of ray-tracing, by solving a certain number of finite directions of radiation, in each cell. North-South, East-West and so on. As I understand it, this means that you can model an infinite amount of reflection, also the model is conservative of energy, whereas I think ray-tracing only tends toward being conservative with enough rays. Is there a similar model available in STAR or planned? Has anyone any experience of radiation modelling in STAR? Also, is it true that Parallel Processing is a pain in STAR?

Re: Radiation in STAR
 

Dear Fluent user,

You should have a look at CFX. The parillelization is excellent and the unstructured, coupled multi-grid solver is also very robust, accurate, and scales linearly (ie 4x grid size takes 4 times longer, not 4^2!).

Regards, Robin

Re: Radiation in STAR
 

The beam tracking scheme in STAR-CD is always conservative, not just with large numbers of beams; (I would like to know the source of this rumor!) In particular, the view factors add up to 1.0, so that all the energy released is accounted for. Increasing the number of beams will improve the accuracy, ie the energy will reach the "correct" (usually smaller) patches which might be missed if the beam discretization is too coarse.

The discrete ordinate method does have certain advantages under some conditions, and it will be offered in a future version of STAR. However, it will need excessive computational resources for basic wall-to-wall radiation, as all the ordinate equations will have to be solved at each iteration (or time step). Increasing the resolution i.e. having more ordinate directions will result in still greater penality per iteration. In contrast, in the beam tracking (or Discrete Transfer) approach, you spend cpu time "upfront" to compute the view factors, and each computational iteration proceeds much faster, with minimal overhead for radiation computations (independantly of the number of beams i.e. beam resolution).

Hope this helps; if you need further information, I would encourage you to contact us directly (agency@adapco.com) for in deapth discussions.

Re: Radiation in STAR
 

As a Fluent user I would like to hear from STAR users what they think about the code? Is it robust general speaking?

Pop

Re: Radiation in STAR
 

Hi

Fluent is so robust that it keeps converging to the same solution even when the boundary conditions have changed drastically. My experience of Fluent solutions bear very little relation to reality but they are 'converged'. I have been using Fluent for about three years. Perhaps they're using too much numerical viscosity?

On the other hand STAR (5 years experience) appears to less likey to have the same ease of convergence for such varying boundary conditions, but funnily enough the solutions appear closer to experimental results.

I just can't understand it!!!