[Jade M]

Why is it difficult for ANSYS to calculate view factors that are small or close to zero? I understand from ANSYS third-party support that this is the case.

I was looking at the theory a bit at the following sites
https://www.sharcnet.ca/Software/Ans...thy_heat5.html
https://www.sharcnet.ca/Software/Ans...lcviewfacmeths
However, I could not glean any information as to why calculating view factors should be so time intensive.

Thanks so much for any information.

[mprinkey]

Originally Posted by Jade M

Why is it difficult for ANSYS to calculate view factors that are small or close to zero? I understand from ANSYS third-party support that this is the case.

I was looking at the theory a bit at the following sites
https://www.sharcnet.ca/Software/Ans...thy_heat5.html
https://www.sharcnet.ca/Software/Ans...lcviewfacmeths
However, I could not glean any information as to why calculating view factors should be so time intensive.

Thanks so much for any information.

The problem is that view factors represent an O(N^2) operation where N is the number of participating surfaces. The Radiative Energy conservation equation is an integral equation spanning the entirety of the boundary of the CFD problem, so it could involve connections from each boundary face to EVERY OTHER boundary face.. If the medium is participating, it then it could involve connections from every finite volume to every other.

The view factor algorithms try to cheat a bit and agglomerate faces that are close and nearly parallel to reduce the size of N. But if there are big temperature or material variations on those surfaces, it may not be possible glue them together. I don't know the algorithm(s) ANSYS is using to simplify the view factor computation and what snags it may run into. But the expense will pile up quickly if the number of participating surfaces become too high. In general, the radiative energy equation leads to a full matrix (versus sparse matrices for almost everything else in CFD). That full matrix not only takes up a lot of memory, but is very costly to solve. Again, I never worked on/with the view factor radiation model in ANSYS, so I can't speak to any pathological cases that you might encounter. Just know that radiation problems (and integral equations in general) are dramatically more difficult to solve than PDEs. That is why we usually rush to turn radiation problems in a single PDE (P1 model) or a relatively small direction set of PDEs (DO model) to try to manage that memory usage and computational complexity. And even with those simplifications, the DO model is often the dominate part of a radiation-coupled CFD problem.

[Jade M]

Thank you for your detailed and timely response. This is very helpful!!